Usage Policy

Legal Rights & Agreement Rules · Context

This saved copy is the September 15, 2025 Usage Policy, replaced by a newer version

The page says this document has been replaced by a newer version linked on the same page and is kept for archival purposes. The copy states it took effect on September 15, 2025, so the supplied text is that dated version rather than the version now published. Neither the replacement text nor its effective date is contained in this source, so no rule now in force can be stated from it.

Supporting quotations

This document has been replaced by a newer version, and is available for archival purposes

Effective September 15, 2025

Qualifications & exceptions

• The replacement notice and the effective date sit above the policy text; the newer version is reachable only through a link.

Not stated here

• The replacement version's text and effective date are not supplied here, so this source cannot state the rules currently in force.
• The policy does not say whether this archived version still binds any user.

Account & Service Control · Context

The policy binds anyone who can submit inputs, including reseller and passthrough access

Anthropic's Usage Policy, also called the Acceptable Use Policy or AUP, applies to anyone who can submit inputs to Anthropic's products and/or services, including through any authorized resellers or passthrough access; the policy calls all of them users. Its stated aim is to help users stay safe and promote responsible use of the products and services. Coverage therefore follows the ability to submit inputs, not the holding of a consumer account or a particular purchase route.

Supporting quotations

applies to anyone who can submit inputs to Anthropic’s products and/or services, including via any authorized resellers or passthrough access, all of whom we refer to as “users.”

The Usage Policy is intended to help our users stay safe and promote the responsible use of our products and services.

Qualifications & exceptions

• Includes access obtained through authorized resellers or passthrough access.

Not stated here

• The policy does not say how it applies to a person whose only access is through a third-party product that does not pass inputs to Anthropic.

Legal Rights & Agreement Rules · Context

The policy has three standards: universal, high-risk use cases, and additional guidelines

The policy is organised by who can use the products and for what purposes. Universal Usage Standards apply to all users and use cases. High-Risk Use Case Requirements apply to specific consumer-facing use cases that pose an elevated risk of harm. Additional Use Case Guidelines apply to certain other use cases, named there as consumer-facing chatbots, products serving minors, agentic use, and Model Context Protocol servers. Anthropic also says it will update the policy as its technology and the associated risks evolve.

Supporting quotations

The Usage Policy is categorized according to who can use our products and for what purposes.

Universal Usage Standards: Our Universal Usage Standards apply to all users and use cases.

High-Risk Use Case Requirements: Our High-Risk Use Case Requirements apply to specific consumer-facing use cases that pose an elevated risk of harm.

Additional Use Case Guidelines: Our Additional Use Case Guidelines apply to certain other use cases, including consumer-facing chatbots, products serving minors, agentic use, and Model Context Protocol servers.

We will update our policy as our technology and the associated risks evolve or as we learn about unanticipated risks.

Qualifications & exceptions

• Each category's detailed rules appear in later sections of the same document; this card states how the document is organised.
• The high-risk category is limited to specific consumer-facing use cases that pose an elevated risk of harm.

Not stated here

• How a use case is judged to pose an elevated risk of harm is not stated in this introduction.

Cross-references

• qualified by (model-proposed): Anthropic may tailor use restrictions for certain government customers by contract — The policy says its Universal Usage Standards apply to all users and use cases and separately that Anthropic may contract with certain governmental customers to tailor use restrictions to that customer’s public mission and legal authorities. The tailoring route adjusts those general standards for named government customers, on Anthropic’s judgment that the tailored restrictions and safeguards adequately mitigate the potential harms.

Account & Service Control · Company practice

Anthropic may throttle, suspend or terminate access and may block or modify outputs

Anthropic's Safeguards Team implements detection and monitoring to enforce the policy, and the policy asks users to review it carefully before using the products or services. If Anthropic learns that a user has violated the Usage Policy, it may throttle, suspend, or terminate that user's access to the products and services. It may also block or modify model outputs when inputs violate the policy. The verbs are discretionary: the policy states what Anthropic may do, not a process it commits to follow, and it names no notice, appeal or reinstatement step.

Supporting quotations

Anthropic’s Safeguards Team will implement detection and monitoring to enforce our Usage Policy

so please review this policy carefully before using our products or services.

If we learn that you have violated our Usage Policy, we may throttle, suspend, or terminate your access to our products and services.

We may also block or modify model outputs when inputs violate our Usage Policy.

Qualifications & exceptions

• Enforcement follows a violation Anthropic learns of; the policy does not say how it learns of one beyond detection and monitoring.

Not stated here

• The policy states no notice, appeal, reinstatement or refund position for a throttled, suspended or terminated account.
• The monitoring methods and thresholds the Safeguards Team uses are not stated.

Security & Safeguards · Procedure

Users can report inaccurate, biased or harmful outputs by email or in-product feedback

If a user believes model outputs are potentially inaccurate, biased or harmful, the policy asks the user to notify Anthropic at usersafety@anthropic.com or to report it directly in the product through the “report issues” thumbs down button or similar feedback features. The in-product route is limited to where those features are available.

Supporting quotations

If you believe that our model outputs are potentially inaccurate, biased or harmful, please notify us at usersafety@anthropic.com, or report it directly in our product through the “report issues” thumbs down button or similar feedback features (where available).

Qualifications & exceptions

• The in-product reporting route exists only where the thumbs down or similar feedback feature is available; the email address is the stated alternative.

Not stated here

• The policy does not say what Anthropic does with a report or how quickly it responds.

Legal Rights & Agreement Rules · Restriction or exception

Anthropic may tailor use restrictions for certain government customers by contract

Anthropic says the policy balances enabling beneficial uses against mitigating potential harms, and that it may enter into contracts with certain governmental customers that tailor use restrictions to that customer's public mission and legal authorities. The tailoring is conditional on Anthropic's judgment that the contractual use restrictions and applicable safeguards adequately mitigate the potential harms addressed by the policy. This is a discretionary route for named government customers, not a general exception available to other users.

Supporting quotations

Anthropic may enter into contracts with certain governmental customers that tailor use restrictions to that customer’s public mission and legal authorities if, in Anthropic’s judgment, the contractual use restrictions and applicable safeguards are adequate to mitigate the potential harms addressed by this Usage Policy.

Qualifications & exceptions

• Conditioned on Anthropic's own judgment that the tailored restrictions and safeguards are adequate to mitigate the potential harms addressed by the Usage Policy.
• Limited to certain governmental customers and their public mission and legal authorities.

Not stated here

• The tailored government contract terms are not supplied and may differ from the rules in this policy.

Cross-references

• qualifies (model-proposed): The policy has three standards: universal, high-risk use cases, and additional guidelines — The policy says its Universal Usage Standards apply to all users and use cases and separately that Anthropic may contract with certain governmental customers to tailor use restrictions to that customer’s public mission and legal authorities. The tailoring route adjusts those general standards for named government customers, on Anthropic’s judgment that the tailored restrictions and safeguards adequately mitigate the potential harms.

Legal Rights & Agreement Rules · Restriction or exception

Users may not use the services to break the law, trade in drugs or facilitate trafficking

Under the Universal Usage Standards, users may not violate applicable laws or engage in illegal activity. The examples given for using the products or services this way are acquiring or exchanging illegal or controlled substances, engaging in or facilitating human trafficking or prostitution, and violating any other applicable laws or regulations in the user's jurisdiction. The examples are introduced by “This includes”, so they illustrate the prohibition rather than exhaust it, and the catch-all points to the law of the user's own jurisdiction.

Supporting quotations

Universal Usage Standards

Do Not Violate Applicable Laws or Engage in Illegal Activity

This includes using our products or services to:

Acquire or exchange illegal or controlled substances

Engage in or facilitate human trafficking or prostitution

Violate any other applicable laws or regulations in your jurisdiction

Qualifications & exceptions

• The listed conduct follows “This includes using our products or services to:”, so the list is illustrative.
• The catch-all limb refers to applicable laws or regulations in the user's jurisdiction.

Not stated here

• The policy does not state which jurisdiction's law governs a user who accesses the services from more than one place.

Content & IP Rights · Restriction or exception

Users may not use the services to infringe a third party's intellectual property rights

The Universal Usage Standards also prohibit using the products or services to infringe, misappropriate, or violate the intellectual property rights of a third party. The prohibition covers infringement and misappropriation of a third party's rights; neither the rights nor the governing law are defined in the policy text.

Supporting quotations

Infringe, misappropriate, or violate the intellectual property rights of a third party

This includes using our products or services to:

Qualifications & exceptions

• The protected interest belongs to a third party, not to the user or to Anthropic.

Not stated here

• The policy does not name the intellectual property rights or the law it relies on.

Security & Safeguards · Restriction or exception

Users may not attack critical infrastructure, critical systems or military infrastructure

Users may not compromise critical infrastructure. The examples are facilitating the destruction or disruption of critical infrastructure such as power grids, water treatment facilities, medical devices, telecommunication networks, or air traffic control systems; obtaining unauthorized access to critical systems such as voting machines, healthcare databases, and financial markets; and interfering with the operation of military bases and related infrastructure. Both “critical infrastructure” and “critical systems” are illustrated by examples rather than defined, so the named systems are examples of a wider class.

Supporting quotations

Do Not Compromise Critical Infrastructure

This includes using our products or services to:

Facilitate the destruction or disruption of critical infrastructure such as power grids, water treatment facilities, medical devices, telecommunication networks, or air traffic control systems

Obtain unauthorized access to critical systems such as voting machines, healthcare databases, and financial markets

Interfere with the operation of military bases and related infrastructure

Qualifications & exceptions

• Critical infrastructure and critical systems are illustrated by examples; the policy does not define their full extent.
• The access limb is limited to unauthorized access.

Not stated here

• The policy does not define which systems count as critical beyond the examples given.

Security & Safeguards · Restriction or exception

Users may not exploit vulnerabilities or gain unauthorized system access

Under the prohibition on compromising computer or network systems, users may not discover or exploit vulnerabilities in systems, networks, or applications without the authorization of the system owner, and may not gain unauthorized access to systems, networks, applications, or devices through technical attacks or social engineering. Authorization is the system owner's, and the policy states no separate route for authorized security testing.

Supporting quotations

Do Not Compromise Computer or Network Systems

This includes using our products or services to:

Discover or exploit vulnerabilities in systems, networks, or applications without authorization of the system owner

Gain unauthorized access to systems, networks, applications, or devices through technical attacks or social engineering

Qualifications & exceptions

• Vulnerability work is prohibited only without the system owner's authorization; the policy does not describe how authorization is obtained.

Not stated here

• The policy does not say whether authorized security testing or disclosure is permitted or on what terms.

Security & Safeguards · Restriction or exception

Users may not create malware, denial-of-service tooling or at-scale compromise tools

The system-security section prohibits creating or distributing malware, ransomware, or other types of malicious code; developing tools for denial-of-service attacks or managing botnets; and creating automated tools designed to compromise multiple systems at scale for malicious purposes. The last example is limited to tools designed for compromise at scale for malicious purposes.

Supporting quotations

Create or distribute malware, ransomware, or other types of malicious code

Develop tools for denial-of-service attacks or managing botnets

Create automated tools designed to compromise multiple systems at scale for malicious purposes

Qualifications & exceptions

• The at-scale tooling limb is limited to tools designed to compromise multiple systems at scale for malicious purposes.

Security & Safeguards · Restriction or exception

Users may not build interception, persistent-access or security-control bypass tools

The same section prohibits creating tools designed to intercept communications or monitor devices without the system owner's authorization; developing persistent access tools designed to operate below normal system security levels, including firmware modifications or hardware implants; and bypassing security controls such as authenticated systems, endpoint protection, or monitoring tools. The interception limb carries the system-owner authorization condition; the persistence and bypass limbs do not repeat it.

Supporting quotations

Create tools designed to intercept communications or monitor devices without authorization of the system owner

Develop persistent access tools designed to operate below normal system security levels, including firmware modifications or hardware implants

Bypass security controls such as authenticated systems, endpoint protection, or monitoring tools

Qualifications & exceptions

• The interception and monitoring limb applies without the system owner's authorization.
• The persistence limb covers tools designed to operate below normal system security levels, including firmware modifications and hardware implants.

Security & Safeguards · Restriction or exception

Users may not design, develop or illegally acquire weapons and their delivery processes

The weapons section prohibits producing, modifying, designing, or illegally acquiring weapons, explosives, dangerous materials, or other systems designed to cause harm to or loss of human life; designing or developing weaponization and delivery processes for the deployment of weapons; and circumventing regulatory controls to acquire weapons or their precursors. Acquisition is prohibited only where illegal, while production, modification and design carry no such qualifier.

Supporting quotations

Do Not Develop or Design Weapons

This includes using our products or services to:

Produce, modify, design, or illegally acquire weapons, explosives, dangerous materials or other systems designed to cause harm to or loss of human life

Design or develop weaponization and delivery processes for the deployment of weapons

Circumvent regulatory controls to acquire weapons or their precursors

Qualifications & exceptions

• Acquisition is limited to illegal acquisition; production, modification and design carry no such qualifier.
• The harm limb covers systems designed to cause harm to or loss of human life.

Not stated here

• The policy does not define weapons, explosives or dangerous materials beyond these descriptions.

Security & Safeguards · Restriction or exception

Users may not develop high-yield explosives or CBRN weapons or their precursors

The weapons section separately prohibits synthesizing or otherwise developing high-yield explosives or biological, chemical, radiological, or nuclear weapons or their precursors, including modifications to evade detection or medical countermeasures. The covered object is the weapon or its precursor, and the expressly covered modification is one made to evade detection or medical countermeasures.

Supporting quotations

Synthesize, or otherwise develop, high-yield explosives or biological, chemical, radiological, or nuclear weapons or their precursors, including modifications to evade detection or medical countermeasures

Qualifications & exceptions

• Covers precursors as well as the weapons, and modifications made to evade detection or medical countermeasures.

Legal Rights & Agreement Rules · Restriction or exception

Users may not incite violence, terrorism or hateful behavior, or discriminate by protected trait

Within the universal prohibitions, users may not incite, facilitate, or promote violent extremism, terrorism, or hateful behavior; provide material support for organizations or individuals associated with violent extremism, terrorism, or hateful behavior; facilitate or promote any act of violence or intimidation targeting individuals, groups, animals, or property; or promote discriminatory practices or behaviors against individuals or groups on the basis of one or more protected attributes such as race, ethnicity, religion, national origin, gender, sexual orientation, or any other identifying trait. The protected attributes are examples introduced by “such as” and closed with “or any other identifying trait”, so the class is not limited to the named traits, and the violence limb reaches animals and property as well as people.

Supporting quotations

Do Not Incite Violence or Hateful Behavior

This includes using our products or services to:

Incite, facilitate, or promote violent extremism, terrorism, or hateful behavior

Provide material support for organizations or individuals associated with violent extremism, terrorism, or hateful behavior

Facilitate or promote any act of violence or intimidation targeting individuals, groups, animals, or property

Promote discriminatory practices or behaviors against individuals or groups on the basis of one or more protected attributes such as race, ethnicity, religion, national origin, gender, sexual orientation, or any other identifying trait

Qualifications & exceptions

• The protected-attribute list is illustrative (“such as”) and ends with “or any other identifying trait”.
• The violence and intimidation limb covers individuals, groups, animals and property.
• Material support is limited to organizations or individuals associated with violent extremism, terrorism or hateful behavior.

Not stated here

• The policy does not define “hateful behavior” or “violent extremism”.

Data Access & Collection · Restriction or exception

Users may not violate privacy laws or collect private, biometric or neural data without permission

Users may not violate privacy rights as defined by applicable privacy laws, such as sharing personal information without consent or accessing private data unlawfully. The same section prohibits misuse, collection, solicitation, or access without permission to private information such as non-public contact details, health data, biometric or neural data (including facial recognition), or confidential or proprietary data. The first limb measures privacy by applicable privacy laws; the second turns on the absence of permission and names biometric and neural data among its examples.

Supporting quotations

Do Not Compromise Privacy or Identity Rights

This includes using our products or services to:

Violate privacy rights as defined by applicable privacy laws, such as sharing personal information without consent or accessing private data unlawfully

Misuse, collect, solicit, or gain access without permission to private information such as non-public contact details, health data, biometric or neural data (including facial recognition), or confidential or proprietary data

Qualifications & exceptions

• Privacy rights are defined by reference to applicable privacy laws, and the examples of violation follow “such as”.
• The misuse, collection, solicitation and access limb is qualified by “without permission”.
• Biometric and neural data, expressly including facial recognition, count as private information.

Not stated here

• The policy does not name the applicable privacy laws or say which law governs a user who is not in the United States.

Legal Rights & Agreement Rules · Restriction or exception

Users may not present results as human-generated or convince people they talk to a human

Users may not impersonate a human by presenting results as human-generated, or use results in a manner intended to convince a natural person that they are communicating with a natural person when they are not. The rule has two limbs: presenting results as human-generated, and using results with that intent. It is framed around what the user does with results, not around naming a particular real person.

Supporting quotations

Impersonate a human by presenting results as human-generated, or using results in a manner intended to convince a natural person that they are communicating with a natural person when they are not

Qualifications & exceptions

• Two limbs: presenting results as human-generated, and using results with the intent to convince a natural person they are communicating with a natural person.

Not stated here

• The policy does not say what disclosure a product built on the services must give to avoid the second limb.

Content & IP Rights · Restriction or exception

Users may not create CSAM or exploit, groom or sexualize minors, including via roleplay

Users may not create, distribute, or promote child sexual abuse material (“CSAM”), including AI-generated CSAM; facilitate the trafficking, sextortion, or any other form of exploitation of a minor; facilitate minor grooming, including generating content designed to impersonate a minor; facilitate child abuse of any form, including instructions for how to conceal abuse; promote or facilitate pedophilic relationships, including via roleplay with the model; or fetishize or sexualize minors, including in fictional settings or via roleplay with the model. The grooming, pedophilic-relationship and sexualization limbs expressly reach roleplay with the model, and the sexualization limb expressly reaches fictional settings.

Supporting quotations

Do Not Compromise Children’s Safety

This includes using our products or services to:

Create, distribute, or promote child sexual abuse material (“CSAM”), including AI-generated CSAM

Facilitate the trafficking, sextortion, or any other form of exploitation of a minor

Facilitate minor grooming, including generating content designed to impersonate a minor

Facilitate child abuse of any form, including instructions for how to conceal abuse

Promote or facilitate pedophilic relationships, including via roleplay with the model

Fetishize or sexualize minors, including in fictional settings or via roleplay with the model

Qualifications & exceptions

• AI-generated CSAM is expressly covered alongside other CSAM.
• The grooming, pedophilic-relationship and sexualization limbs expressly reach roleplay with the model.
• The sexualization prohibition applies in fictional settings as well.
• Child abuse instructions extend to how to conceal abuse.

Cross-references

• defined by (model-proposed): A minor is anyone under 18 regardless of jurisdiction, and detected CSAM is reported — The children’s safety section defines a minor or child as anyone under 18 regardless of jurisdiction and describes reporting on detection of CSAM or coercion or enticement of a minor. That definition fixes the age scope of the CSAM, grooming and sexualization prohibitions in the same section.

Legal Rights & Agreement Rules · Company practice

A minor is anyone under 18 regardless of jurisdiction, and detected CSAM is reported

The policy defines a minor or child as any individual under the age of 18 years old, regardless of jurisdiction. Where Anthropic detects CSAM (including AI-generated CSAM), or coercion or enticement of a minor to engage in sexual activities, it will report to relevant authorities. The age definition applies whatever the local age of majority, and the reporting commitment is stated to follow detection of that conduct.

Supporting quotations

Note: We define a minor or child to be any individual under the age of 18 years old, regardless of jurisdiction.

When we detect CSAM (including AI-generated CSAM), or coercion or enticement of a minor to engage in sexual activities, we will report to relevant authorities.

Qualifications & exceptions

• The under-18 definition applies regardless of jurisdiction.
• Reporting follows detection of CSAM or of coercion or enticement of a minor to engage in sexual activities.

Not stated here

• The policy does not name the authorities, a reporting timeline, or what happens to the account.

Cross-references

• defines (model-proposed): Users may not create CSAM or exploit, groom or sexualize minors, including via roleplay — The children’s safety section defines a minor or child as anyone under 18 regardless of jurisdiction and describes reporting on detection of CSAM or coercion or enticement of a minor. That definition fixes the age scope of the CSAM, grooming and sexualization prohibitions in the same section.

Content & IP Rights · Restriction or exception

Users may not generate content promoting suicide, self-harm, bullying, animal cruelty or gore

Users may not create psychologically or emotionally harmful content. The examples are facilitating, promoting, or glamorizing any form of suicide or self-harm, including disordered eating and unhealthy or compulsive exercise; engaging in behaviors that promote unhealthy or unattainable body image or beauty standards, such as using the model to critique anyone’s body shape or size; shaming, humiliating, intimidating, bullying, harassing, or celebrating the suffering of individuals; coordinating harassment or intimidation of an individual or group; generating content depicting animal cruelty or abuse; promoting, trivializing, or depicting graphic violence or gratuitous gore, including sexual violence; or developing a new product or service, or supporting an existing one, that employs or facilitates deceptive techniques with the intent of causing emotional harm. The body-image limb names critiquing a person’s body shape or size as an example, and the product limb is limited to deceptive techniques used with intent to cause emotional harm.

Supporting quotations

Do Not Create Psychologically or Emotionally Harmful Content

This includes using our products or services to:

Facilitate, promote, or glamorize any form of suicide or self-harm, including disordered eating and unhealthy or compulsive exercise

Engage in behaviors that promote unhealthy or unattainable body image or beauty standards, such as using the model to critique anyone’s body shape or size

Shame, humiliate, intimidate, bully, harass, or celebrate the suffering of individuals

Coordinate the harassment or intimidation of an individual or group

Generate content depicting animal cruelty or abuse

Promote, trivialize, or depict graphic violence or gratuitous gore, including sexual violence

Develop a new product or service, or support an existing product or service that employs or facilitates deceptive techniques with the intent of causing emotional harm

Qualifications & exceptions

• The body-image limb gives using the model to critique anyone’s body shape or size as an example.
• The product-development limb is limited to employing or facilitating deceptive techniques with the intent of causing emotional harm.
• The violence limb expressly includes sexual violence, graphic violence and gratuitous gore.

Not stated here

• The policy does not define “graphic violence” or “gratuitous gore”.

Legal Rights & Agreement Rules · Restriction or exception

Users may not create deceptive information, fake personas or false health and science claims

Users may not create or disseminate deceptive or misleading information about, or with the intention of targeting, a group, entity or person; deceptive or misleading information about laws, regulations, procedures, practices, or standards established by an institution, entity or governing body; conspiratorial narratives meant to target a specific group, individual or entity; impersonation of real entities, or fake personas, to falsely attribute content or mislead others about its origin without consent or legal right; or false or misleading information related to medical, health or science issues. The first limb reaches information about a group, entity or person or created with the intention of targeting one, and the entity-impersonation limb is qualified by “without consent or legal right”.

Supporting quotations

Do Not Create or Spread Misinformation

This includes using our products or services to:

Create or disseminate deceptive or misleading information about, or with the intention of targeting, a group, entity or person

Create or disseminate deceptive or misleading information about laws, regulations, procedures, practices, standards established by an institution, entity or governing body

Create or disseminate conspiratorial narratives meant to target a specific group, individual or entity

Impersonate real entities or create fake personas to falsely attribute content or mislead others about its origin without consent or legal right

Provide false or misleading information related to medical, health or science issues

Qualifications & exceptions

• The first limb covers information about a group, entity or person, or information created with the intention of targeting one.
• Impersonating real entities or creating fake personas is prohibited without consent or legal right.
• The medical, health or science limb is stated without a targeting requirement.

Not stated here

• The policy does not define how falsity is assessed or who assesses it.

Legal Rights & Agreement Rules · Restriction or exception

Users may not target voters, run deceptive political movements or disrupt elections

Users may not undermine democratic processes or engage in targeted campaign activities. The listed examples are personalized vote or campaign targeting based on individual profiles or data; artificial or deceptive political movements in which the source, scale or nature of the campaign or activities is misrepresented; automated communications to public officials or voters at scale that conceal their artificial origin, or systematic vote solicitation that could undermine election integrity; political content designed to deceive or mislead voters, including synthetic media of political figures; false or misleading information in political and electoral contexts, including about candidates, parties, policies, voting procedures or election security; political lobbying or grassroots advocacy using false or fabricated information, or lobbying or advocacy materials containing demonstrably false claims about facts, data or events; inciting, glorifying or facilitating disruption of electoral or civic processes, including interference with voting systems, vote counting or certification; and content designed to suppress voter turnout or discourage legitimate political participation through deception or intimidation.

Supporting quotations

Do Not Undermine Democratic Processes or Engage in Targeted Campaign Activities

This includes using our products or services to:

Engage in personalized vote or campaign targeting based on individual profiles or data

Create artificial or deceptive political movements in which the source, scale or nature of the campaign or activities is misrepresented

Generate automated communications to public officials or voters at scale that conceal their artificial origin, or engage in systematic vote solicitation that could undermine election integrity

Create political content designed to deceive or mislead voters, including synthetic media of political figures

Generate or disseminate false or misleading information in political and electoral contexts, including about candidates, parties, policies, voting procedures, or election security

Engage in political lobbying or grassroots advocacy using false or fabricated information, or create lobbying or advocacy materials containing demonstrably false claims about facts, data, or events

Incite, glorify or facilitate the disruption of electoral or civic processes, including interference with voting systems, vote counting, or certification processes

Create content designed to suppress voter turnout or discourage legitimate political participation through deception or intimidation

Qualifications & exceptions

• The automated-communications limb is limited to messages that conceal their artificial origin, or to systematic vote solicitation that could undermine election integrity.
• The lobbying limb reaches advocacy using false or fabricated information and materials containing demonstrably false claims about facts, data or events.
• Several limbs turn on deception or intimidation rather than on political content as such.

Not stated here

• The policy does not define “election integrity” or say who assesses whether conduct could undermine it.

Legal Rights & Agreement Rules · Restriction or exception

Users may not make criminal justice determinations or use models to impair legal rights

Users may not make determinations on criminal justice applications, including making decisions about or determining eligibility for parole or sentencing. The same section prohibits utilizing models as part of any law enforcement application that violates or impairs the liberty, civil liberties, or human rights of natural persons. The first limb names parole and sentencing determinations as examples of criminal justice applications; the second is framed by the effect of the application on liberty, civil liberties or human rights.

Supporting quotations

Do Not Use for Criminal Justice, Censorship, Surveillance, or Prohibited Law Enforcement Purposes

This includes using our products or services to:

Make determinations on criminal justice applications, including making decisions about or determining eligibility for parole or sentencing

Utilize models as part of any law enforcement application that violates or impairs the liberty, civil liberties, or human rights of natural persons

Qualifications & exceptions

• Parole and sentencing are given as examples of criminal justice determinations (“including”).
• The law-enforcement limb is limited to applications that violate or impair liberty, civil liberties or human rights.

Not stated here

• The policy does not define “criminal justice applications” beyond the parole and sentencing examples.

Legal Rights & Agreement Rules · Restriction or exception

Users may not track location, emotion or communication without consent, or do facial recognition

Users may not target or track a person’s physical location, emotional state, or communication without their consent, including using the products for facial recognition, battlefield management applications or predictive policing. The prohibition is qualified by the absence of the person’s consent, and facial recognition, battlefield management and predictive policing are given as examples of the prohibited uses.

Supporting quotations

Target or track a person’s physical location, emotional state, or communication without their consent, including using our products for facial recognition, battlefield management applications or predictive policing

Qualifications & exceptions

• The tracking prohibition applies without the person’s consent.
• Facial recognition, battlefield management applications and predictive policing are examples (“including”), not a closed list.

Not stated here

• The policy does not say how consent must be obtained or evidenced.

Cross-references

• contrast (model-proposed): Users may not build emotion recognition systems except for medical or safety reasons — One prohibition bars targeting or tracking a person’s emotional state without their consent; another bars building or supporting emotion recognition systems used to infer emotions, except for medical or safety reasons. Consent removes the first prohibition but is not a stated route around the second, so a consented emotion-inference system still needs a medical or safety purpose. The two provisions therefore condition emotion inference on different tests.

Inference & Profiling · Restriction or exception

Users may not score people's trustworthiness or categorize them by sensitive attributes

Users may not utilize models to assign scores or ratings to individuals based on an assessment of their trustworthiness or social behavior without notification or their consent. The same section prohibits utilizing models as part of any biometric categorization system for categorizing people based on their biometric data to infer their race, political opinions, trade union membership, religious or philosophical beliefs, sex life or sexual orientation. The scoring limb is qualified by the absence of notification or consent, and the categorization limb names the sensitive attributes it may not be used to infer.

Supporting quotations

Utilize models to assign scores or ratings to individuals based on an assessment of their trustworthiness or social behavior without notification or their consent

Utilize models as part of any biometric categorization system for categorizing people based on their biometric data to infer their race, political opinions, trade union membership, religious or philosophical beliefs, sex life or sexual orientation

Qualifications & exceptions

• Scoring or rating trustworthiness or social behavior is prohibited without notification or consent.
• The biometric categorization limb covers inference of race, political opinions, trade union membership, religious or philosophical beliefs, sex life or sexual orientation.

Not stated here

• The policy does not say what notice satisfies the notification requirement.

Inference & Profiling · Restriction or exception

Users may not build emotion recognition systems except for medical or safety reasons

Users may not build or support emotional recognition systems or techniques that are used to infer emotions of a natural person, except for medical or safety reasons. The exception is stated by its purpose, so the prohibition applies to emotion-inference systems that are not built or supported for medical or safety reasons.

Supporting quotations

Build or support emotional recognition systems or techniques that are used to infer emotions of a natural person, except for medical or safety reasons

Qualifications & exceptions

• The exception is limited to medical or safety reasons.

Not stated here

• The policy does not say who determines whether a purpose is medical or safety-related.

Cross-references

• contrast (model-proposed): Users may not track location, emotion or communication without consent, or do facial recognition — One prohibition bars targeting or tracking a person’s emotional state without their consent; another bars building or supporting emotion recognition systems used to infer emotions, except for medical or safety reasons. Consent removes the first prohibition but is not a stated route around the second, so a consented emotion-inference system still needs a medical or safety purpose. The two provisions therefore condition emotion inference on different tests.

Legal Rights & Agreement Rules · Restriction or exception

Users may not analyze or identify content to censor for a government organization

Users may not analyze or identify specific content to censor on behalf of a government organization. The prohibition turns on the purpose of the analysis and the identity of the beneficiary, not on the content analyzed.

Supporting quotations

Analyze or identify specific content to censor on behalf of a government organization

Not stated here

• The policy does not define “government organization” or address work for private intermediaries.

Legal Rights & Agreement Rules · Restriction or exception

Users may not facilitate production or distribution of counterfeit or illicitly acquired goods

Users may not facilitate the production, acquisition, or distribution of counterfeit or illicitly acquired goods. The prohibition covers production, acquisition and distribution, and applies to goods that are counterfeit or were acquired illicitly.

Supporting quotations

Do Not Engage in Fraudulent, Abusive, or Predatory Practices

This includes using our products or services to:

Facilitate the production, acquisition, or distribution of counterfeit or illicitly acquired goods

Not stated here

• The policy does not define “illicitly acquired goods” or list covered product categories.

Legal Rights & Agreement Rules · Restriction or exception

Users may not promote or facilitate the generation or distribution of spam

Users may not promote or facilitate the generation or distribution of spam. The prohibition reaches both generating and distributing spam, and promoting or facilitating those activities by others.

Supporting quotations

Promote or facilitate the generation or distribution of spam

Not stated here

• The policy does not define spam or set volume thresholds.

Charges & Financial Obligations · Restriction or exception

Users may not generate content for frauds, scams, phishing or malware causing harm

Users may not generate content for fraudulent activities, schemes, scams, phishing, or malware that can result in direct financial or psychological harm. The harm qualifier attaches to the listed activities, and the prohibition is framed around generating content for them rather than around carrying them out.

Supporting quotations

Generate content for fraudulent activities, schemes, scams, phishing, or malware that can result in direct financial or psychological harm

Qualifications & exceptions

• The listed activities are qualified by “that can result in direct financial or psychological harm”.

Not stated here

• The policy does not define “direct” financial or psychological harm.

Legal Rights & Agreement Rules · Restriction or exception

Users may not create falsified documents such as fake IDs, licences or currency

Users may not create falsified documents, including fake IDs, licenses, currency, or other government documents. The listed items are examples introduced by “including”, and currency and other government documents fall within the same prohibition as identity and licence documents.

Supporting quotations

Create falsified documents including fake IDs, licenses, currency, or other government documents

Qualifications & exceptions

• Fake IDs, licenses, currency and other government documents are examples, not a closed list.

Content & IP Rights · Restriction or exception

Users may not build fraudulent products or generate fake reviews, comments or media

Users may not develop, promote, or otherwise facilitate the sale or distribution of fraudulent or deceptive products, or generate deceptive or misleading digital content such as fake reviews, comments, or media. The first limb covers development, promotion and facilitation of sale or distribution; the second covers generated digital content that is deceptive or misleading, with fake reviews, comments and media as examples.

Supporting quotations

Develop, promote, or otherwise facilitate the sale or distribution of fraudulent or deceptive products

Generate deceptive or misleading digital content such as fake reviews, comments, or media

Qualifications & exceptions

• The product limb covers development, promotion and facilitation of sale or distribution, not only manufacturing.
• Fake reviews, comments and media are examples of deceptive digital content.

Not stated here

• The policy does not define what makes a product fraudulent or deceptive.

Charges & Financial Obligations · Restriction or exception

Users may not run pyramid schemes or predatory short-term high-interest lending

Users may not engage in or facilitate multi-level marketing, pyramid schemes, or other deceptive business models that use high-pressure sales tactics or exploit participants. The same section prohibits promoting or facilitating payday loans, title loans, or other high-interest, short-term lending practices that exploit vulnerable individuals. The business-model limb reaches models that use high-pressure sales tactics or exploit participants, and the lending limb is limited to practices that exploit vulnerable individuals.

Supporting quotations

Engage in or facilitate multi-level marketing, pyramid schemes, or other deceptive business models that use high-pressure sales tactics or exploit participants

Promote or facilitate payday loans, title loans, or other high-interest, short-term lending practices that exploit vulnerable individuals

Qualifications & exceptions

• The business-model limb is limited to models that use high-pressure sales tactics or exploit participants.
• The lending limb covers payday, title and other high-interest short-term lending that exploits vulnerable individuals.

Not stated here

• The policy does not set an interest-rate threshold or define “vulnerable individuals”.

Charges & Financial Obligations · Restriction or exception

Users may not exploit people by age, disability or economic situation or harass debtors

Users may not engage in deceptive or abusive practices that exploit individuals based on age, disability or a specific social or economic situation. The same section prohibits promoting or facilitating abusive or harassing debt collection practices. The exploitation limb names the grounds on which it is prohibited; the debt-collection limb covers promotion and facilitation as well as direct practice.

Supporting quotations

Engage in deceptive or abusive practices that exploit individuals based on age, disability or a specific social or economic situation

Promote or facilitate the use of abusive or harassing debt collection practices

Qualifications & exceptions

• The exploitation limb is framed around age, disability or a specific social or economic situation.
• The debt-collection limb reaches promoting or facilitating abusive or harassing practices, not only performing them.

Not stated here

• The policy does not define “abusive” debt collection by reference to any named statute.

Legal Rights & Agreement Rules · Restriction or exception

Users may not deploy subliminal or manipulative techniques that impair decision-making

Users may not develop a product, or support an existing service, that deploys subliminal, manipulative, or deceptive techniques to distort behavior by impairing decision-making. The prohibition applies to building or supporting such a product or service, and the technique must be used to distort behavior by impairing decision-making.

Supporting quotations

Develop a product or support an existing service that deploys subliminal, manipulative, or deceptive techniques to distort behavior by impairing decision-making

Qualifications & exceptions

• The prohibited effect is distorting behavior by impairing decision-making.
• Both developing a product and supporting an existing service are covered.

Not stated here

• The policy does not define “subliminal” or “manipulative” techniques.

Legal Rights & Agreement Rules · Restriction or exception

Users may not circumvent the guardrails or terms of other platforms and services

Users may not engage in actions or behaviors that circumvent the guardrails or terms of other platforms or services. The prohibition concerns conduct directed at third-party platforms and services, not at Anthropic’s own guardrails, which are addressed separately in the platform-abuse section.

Supporting quotations

Engage in actions or behaviors that circumvent the guardrails or terms of other platforms or services

Not stated here

• The policy does not name which third-party terms or guardrails it refers to.

Content & IP Rights · Restriction or exception

Users may not plagiarize or submit AI-assisted work without permission or attribution

Users may not plagiarize or submit AI-assisted work without proper permission or attribution. The prohibition covers both plagiarism and submission of AI-assisted work that lacks proper permission or attribution, so attribution and permission are the two stated conditions.

Supporting quotations

Plagiarize or submit AI-assisted work without proper permission or attribution

Qualifications & exceptions

• Submission of AI-assisted work is prohibited without proper permission or attribution.

Not stated here

• The policy does not define “proper” permission or attribution.

Account & Service Control · Restriction or exception

Users may not coordinate malicious activity across accounts or automate account creation

Users may not coordinate malicious activity across multiple accounts to avoid detection or circumvent product guardrails, or generate identical or similar inputs that otherwise violate the Usage Policy. The same section prohibits utilizing automation in account creation or to engage in spammy behavior. The first limb reaches coordination across accounts and repeated policy-violating inputs; the second reaches automated account creation and spammy automation.

Supporting quotations

Do Not Abuse our Platform

This includes using our products or services to:

Coordinate malicious activity across multiple accounts to avoid detection or circumvent product guardrails or generating identical or similar inputs that otherwise violate our Usage Policy

Utilize automation in account creation or to engage in spammy behavior

Qualifications & exceptions

• Coordination across accounts is prohibited when used to avoid detection or circumvent guardrails, or when it generates identical or similar violating inputs.
• Automation is prohibited in account creation and for spammy behavior.

Not stated here

• The policy does not say how identical or similar inputs are assessed.

Account & Service Control · Restriction or exception

Users may not circumvent a ban through another account or by giving banned parties access

Users may not circumvent a ban through the use of a different account, such as the creation of a new account, use of an existing account, or providing access to a person or entity that was previously banned. The examples show that circumvention includes both switching accounts and passing access to a banned person or entity.

Supporting quotations

Circumvent a ban through the use of a different account, such as the creation of a new account, use of an existing account, or providing access to a person or entity that was previously banned

Qualifications & exceptions

• The examples are new accounts, existing accounts, and providing access to a previously banned person or entity.

Account & Service Control · Restriction or exception

Users may not give Claude access to people in unsupported regions

Users may not access or facilitate account or API access to Claude to persons, entities, or users in violation of the Supported Regions Policy. The prohibition reaches both the user’s own access and facilitating access for others, and its scope is set by reference to the linked Supported Regions Policy.

Supporting quotations

Access or facilitate account or API access to Claude to persons, entities, or users in violation of our Supported Regions Policy

Qualifications & exceptions

• Both accessing and facilitating access for others are covered.
• The governing list is the Supported Regions Policy linked from the source.

Not stated here

• The linked Supported Regions Policy text was not supplied in this source; the list of regions is therefore unknown here.

Security & Safeguards · Restriction or exception

Users may not jailbreak or prompt-inject to bypass guardrails without Anthropic authorization

Users may not intentionally bypass capabilities, restrictions, or guardrails established within the products for the purposes of instructing the model to produce harmful outputs (for example jailbreaking or prompt injection) without prior authorization from Anthropic. The prohibition turns on intent to instruct the model to produce harmful outputs, and prior authorization from Anthropic removes it.

Supporting quotations

Intentionally bypass capabilities, restrictions, or guardrails established within our products for the purposes of instructing the model to produce harmful outputs (e.g., jailbreaking or prompt injection) without prior authorization from Anthropic

Qualifications & exceptions

• Jailbreaking and prompt injection are given as examples.
• The bypass must be for the purpose of instructing the model to produce harmful outputs.
• Prior authorization from Anthropic is the stated exception.

Not stated here

• The policy does not describe how prior authorization is requested or granted.

Content & IP Rights · Restriction or exception

Users may not train an AI model on inputs and outputs without Anthropic authorization

Users may not utilize inputs and outputs to train an AI model (for example “model scraping” or “model distillation”) without prior authorization from Anthropic. The prohibition covers both scraping and distillation as examples, and prior authorization from Anthropic is the stated exception.

Supporting quotations

Utilization of inputs and outputs to train an AI model (e.g., “model scraping” or “model distillation”) without prior authorization from Anthropic

Qualifications & exceptions

• Model scraping and model distillation are examples of training on inputs and outputs.
• Prior authorization from Anthropic is the stated exception.

Not stated here

• The policy does not say what authorization process applies or which uses are approved.

Content & IP Rights · Restriction or exception

Users may not generate sexually explicit content, fetish content or erotic chats

Users may not depict or request sexual intercourse or sex acts, generate content related to sexual fetishes or fantasies, or engage in erotic chats. These items are stated without an age or consent qualifier in this section; content involving minors is addressed separately in the children’s safety section, which prohibits sexualizing minors including in fictional settings or roleplay.

Supporting quotations

Do Not Generate Sexually Explicit Content

This includes using our products or services to:

Depict or request sexual intercourse or sex acts

Generate content related to sexual fetishes or fantasies

Engage in erotic chats

Qualifications & exceptions

• The section states no age or consent qualifier for these items.
• Content involving minors is governed by the separate children’s safety prohibition.

Not stated here

• The policy does not define “sexual fetishes or fantasies” or draw a line for non-explicit sexual content.

Content & IP Rights · Restriction or exception

Users may not facilitate, promote or depict incest or bestiality

Users may not facilitate, promote, or depict incest or bestiality. The prohibition covers facilitating and promoting as well as depicting, and it is stated separately from the other sexually explicit content items.

Supporting quotations

Facilitate, promote, or depict incest or bestiality

Qualifications & exceptions

• Facilitating and promoting are covered, not only depiction.

Legal Rights & Agreement Rules · Restriction or exception

High-risk use cases must implement human review and AI disclosure measures

Anthropic requires users and organizations to implement two additional safety measures for the High-Risk Use Cases described in the section: human-in-the-loop review by a qualified professional and disclosure that AI is being used. The stated rationale is that some use cases pose an elevated risk of harm because they influence domains vital to public welfare and social equity, so relevant human expertise should be integrated and end-users should be aware when AI has been involved in producing outputs. The measures are stated as requirements for those uses, not as guidance.

Supporting quotations

High-Risk Use Case Requirements

Some use cases pose an elevated risk of harm because they influence domains that are vital to public welfare and social equity.

for the “High-Risk Use Cases” described below, we require that you implement these additional safety measures

Qualifications & exceptions

• The requirements attach only to the uses the policy itself labels High-Risk Use Cases.
• Two measures are required: human-in-the-loop review and disclosure.
• The rationale paragraph describes influence on domains vital to public welfare and social equity; the introductory summary of the policy describes the requirements as applying to specific consumer-facing use cases.

Not stated here

• The policy does not say how a user should determine that its use falls in a listed category, or what happens to uses that span a listed and an unlisted domain.

Cross-references

• qualified by (model-proposed): A qualified professional must review high-risk advice before it reaches people — The requirements section states that two additional safety measures apply to the High-Risk Use Cases. The human-in-the-loop measure then specifies its trigger, the required reviewer and the timing of review, so the duty is narrower than every use that appears on the High-Risk Use Cases list.

Legal Rights & Agreement Rules · Restriction or exception

A qualified professional must review high-risk advice before it reaches people

When the products or services are used to provide advice, recommendations, or in subjective decision-making that directly affects individuals or consumers, a qualified professional in that field must review the content or decision before it is disseminated or finalized. The user or organization remains responsible for the accuracy and appropriateness of that information. The trigger is the type of use and its direct effect on individuals or consumers, so the duty is narrower than every use that appears on the High-Risk Use Cases list.

Supporting quotations

When using our products or services to provide advice, recommendations, or in subjective decision-making directly affecting individuals or consumers, a qualified professional in that field must review the content or decision prior to dissemination or finalization.

You or your organization are responsible for the accuracy and appropriateness of that information.

Qualifications & exceptions

• The trigger is providing advice, recommendations, or subjective decision-making that directly affects individuals or consumers.
• Review must occur before dissemination or finalization.
• The reviewer must be a qualified professional in that field.
• The user or organization remains responsible for the accuracy and appropriateness of the information.

Not stated here

• The policy does not define “qualified professional” or describe required credentials.
• The policy does not say whether review is needed for each output or may cover a process.

Cross-references

• qualifies (model-proposed): High-risk use cases must implement human review and AI disclosure measures — The requirements section states that two additional safety measures apply to the High-Risk Use Cases. The human-in-the-loop measure then specifies its trigger, the required reviewer and the timing of review, so the duty is narrower than every use that appears on the High-Risk Use Cases list.

Legal Rights & Agreement Rules · Restriction or exception

Consumers must be told AI helped produce high-risk advice or decisions

If model outputs are presented directly to individuals or consumers, the user must disclose that AI is being used to help produce the advice, decisions, or recommendations. The disclosure must be given at a minimum at the beginning of each session. The duty is triggered by direct presentation of outputs to individuals or consumers, and the stated minimum is a timing floor rather than a limit on how often disclosure may be given.

Supporting quotations

If model outputs are presented directly to individuals or consumers, you must disclose to them that you are using AI to help produce your advice, decisions, or recommendations.

This disclosure must be provided at a minimum at the beginning of each session.

Qualifications & exceptions

• The trigger is model outputs presented directly to individuals or consumers.
• Disclosure must be given at a minimum at the beginning of each session.

Not stated here

• The policy does not prescribe the wording or form of the disclosure.

Legal Rights & Agreement Rules · Restriction or exception

High-risk categories cover legal, insurance, finance, employment, education and media uses

The policy lists the High-Risk Use Cases that trigger the additional measures. Besides healthcare, which is carded separately, they are: legal interpretation, legal guidance, or decisions with legal implications; health, life, property, disability or other insurance underwriting, claims processing or coverage decisions; financial decisions including investment advice, loan approvals and determining financial eligibility or creditworthiness; employment and housing decisions, including employability, resume screening, hiring tools, employment determinations and eligibility for housing including leases and home loans; academic testing, accreditation and admissions, including standardized testing companies that administer school admissions and agencies that evaluate and certify educational institutions; and media or professional journalistic content generated automatically and published for external consumption.

Supporting quotations

“High-Risk Use Cases” include:

Legal: Use cases related to legal interpretation, legal guidance, or decisions with legal implications

Insurance: Use cases related to health, life, property, disability, or other types of insurance underwriting, claims processing, or coverage decisions

Finance: Use cases related to financial decisions, including investment advice, loan approvals, and determining financial eligibility or creditworthiness

Employment and housing: Use cases related to decisions about the employability of individuals, resume screening, hiring tools, or other employment determinations or decisions regarding eligibility for housing, including leases and home loans

Academic testing, accreditation and admissions: Use cases related to standardized testing companies that administer school admissions

Media or professional journalistic content: Use cases related to using our products or services to automatically generate content and publish it for external consumption

Qualifications & exceptions

• The list is introduced as the uses that count as High-Risk Use Cases, so it defines where the human-in-the-loop and disclosure measures attach.
• The media category covers automatic generation of content published for external consumption.
• Employment and housing are grouped in one item.

Not stated here

• The policy does not say whether a use that only partly matches a category is covered.

Cross-references

• narrowed by (model-proposed): Healthcare is a high-risk category, but general wellness advice is excluded — The healthcare item is listed among the High-Risk Use Cases, and the sentence that follows it puts general wellness advice outside that category. Read together, the category covers healthcare decisions and medical guidance but not wellness advice such as sleep, stress, nutrition or exercise.

Legal Rights & Agreement Rules · Restriction or exception

Healthcare is a high-risk category, but general wellness advice is excluded

Healthcare use cases related to healthcare decisions, medical diagnosis, patient care, therapy, mental health, or other medical guidance count as High-Risk Use Cases. Wellness advice, such as advice on sleep, stress, nutrition or exercise, does not fall under this category. The carve-out is stated by example, so it removes general wellness advice from the healthcare category without saying that such advice escapes every other part of the policy.

Supporting quotations

Healthcare: Use cases related to healthcare decisions, medical diagnosis, patient care, therapy, mental health, or other medical guidance.

Wellness advice (e.g., advice on sleep, stress, nutrition, exercise, etc.) does not fall under this category

Qualifications & exceptions

• The item covers healthcare decisions, medical diagnosis, patient care, therapy, mental health and other medical guidance.
• Wellness advice on sleep, stress, nutrition and exercise is expressly outside the healthcare category.
• The carve-out removes the use from this category only; it does not state that other policy obligations are inapplicable.

Not stated here

• The policy does not draw the line between medical guidance and wellness advice where they overlap, such as stress or sleep advice framed therapeutically.

Cross-references

• narrows (model-proposed): High-risk categories cover legal, insurance, finance, employment, education and media uses — The healthcare item is listed among the High-Risk Use Cases, and the sentence that follows it puts general wellness advice outside that category. Read together, the category covers healthcare decisions and medical guidance but not wellness advice such as sleep, stress, nutrition or exercise.

Legal Rights & Agreement Rules · Restriction or exception

Additional use case guidelines apply even when a use is not high-risk

A second set of requirements applies to the uses listed in the Additional Use Case Guidelines section: consumer-facing chatbots, products serving minors, agentic use, and Model Context Protocol servers. Those uses must comply with the guidance regardless of whether they are also High-Risk Use Cases, so the two layers of requirements can apply to the same use at the same time.

Supporting quotations

Additional Use Case Guidelines

regardless of whether they are High-Risk Use Cases

Qualifications & exceptions

• The guidance applies regardless of whether the use is also a High-Risk Use Case.
• The listed uses are consumer-facing chatbots, products serving minors, agentic use and MCP servers.

Not stated here

• The policy does not say how the guidelines interact where a use falls into more than one listed category.

Legal Rights & Agreement Rules · Restriction or exception

All consumer-facing chatbots must say users are talking to AI, not a human

All consumer-facing chatbots, including any external-facing or interactive AI agent, must disclose to users that they are interacting with AI rather than a human. The disclosure must be provided at a minimum at the beginning of each chat session. The duty is stated for every consumer-facing chatbot, without a risk threshold.

Supporting quotations

All consumer-facing chatbots, including any external-facing or interactive AI agent, must disclose to users that they are interacting with AI rather than a human.

This disclosure must be provided at a minimum at the beginning of each chat session.

Qualifications & exceptions

• The duty covers any external-facing or interactive AI agent, not only products described as chatbots.
• The stated minimum timing is the beginning of each chat session.

Not stated here

• The policy does not prescribe wording, and does not say whether a persistent banner satisfies the per-session minimum.

Legal Rights & Agreement Rules · Restriction or exception

Products serving minors must follow Anthropic’s additional minors guidelines

Products serving minors, including organizations that let minors interact directly with products incorporating Anthropic’s APIs, must comply with the additional guidelines in the linked Help Center article on responsible use for organizations serving minors. The policy states the obligation and points to the article for its content.

Supporting quotations

Products serving minors, including organizations providing minors with the ability to directly interact with products that incorporate our API(s), must comply with the additional guidelines outlined in our Help Center article.

Qualifications & exceptions

• The item covers direct interaction by minors with API-based products, as well as products serving minors generally.
• The substantive guidelines are in the linked Help Center article, not in this document.

Not stated here

• The linked Help Center article was not supplied in this source, so its guidelines are unknown here.
• The policy does not define “products serving minors” by reference to an age or share of users.

Legal Rights & Agreement Rules · Restriction or exception

Agentic use cases remain subject to the Usage Policy

Agentic use cases must still comply with the Usage Policy. The section points to a linked Help Center article for examples of how the Usage Policy prohibitions apply in agentic use, so agentic operation does not create an exemption from the prohibitions elsewhere in the policy.

Supporting quotations

Agentic use cases must still comply with the Usage Policy.

Qualifications & exceptions

• The requirement is stated as continuing compliance with the Usage Policy, not as a separate rule set.
• Worked examples of the prohibitions in agentic contexts are in the linked article.

Not stated here

• The linked Help Center article was not supplied in this source, so its agentic examples are unknown here.

Legal Rights & Agreement Rules · Restriction or exception

MCP servers in the Connector Directory must follow the Directory Policy

Model Context Protocol servers listed in Anthropic’s Connector Directory must comply with the linked Directory Policy. The requirement attaches to servers that are listed in the directory and points to that policy for its content.

Supporting quotations

Model Context Protocol (MCP) servers listed in our Connector Directory must comply with our Directory Policy.

Qualifications & exceptions

• The requirement attaches to MCP servers listed in the Connector Directory.
• The substantive rules are in the linked Directory Policy, not in this document.

Not stated here

• The linked Directory Policy was not supplied in this source, so its content is unknown here.