Manus says it uses safety measures to protect users and comply with law, including detection of illegal content
Manus says it implements safety measures to help protect users and to comply with applicable law, including detection mechanisms for illegal content, and that it uses measures to prevent, detect and respond to illegal or harmful content. The wording is that the measures help protect users; it is not a promise that harmful content never appears.
Supporting quotations
We implement safety measures to help protect users and comply with applicable law, including detection mechanisms for illegal content.
We employ safety measures to prevent, detect, and respond to illegal or harmful content.
Qualifications & exceptions
Stated as measures that "help protect" users; no guarantee of outcome.
Not stated here
The policy does not say which measures apply beyond those it lists in later sections.
Manus partly relies on safety controls built into the third-party AI models it uses
In addition to its own measures, Manus relies on safety controls built into the third-party AI models and tools it integrates. Those third parties maintain their own safety controls and content policies, so the safeguards that apply to a task are partly set by outside providers rather than by this policy alone.
Supporting quotations
We also rely on safety controls built into the third-party AI models we use.
We use third-party AI models and tools that maintain their own safety controls and content policies.
Third-party model safeguards: We rely on safety controls built into the AI models we integrate;
Not stated here
The third parties' own safety controls and content policies are not supplied in this document and are not described.
Manus commits to prevent obviously unlawful or high-risk outputs and to detect and act quickly on problematic content
Manus says it balances access to useful information against reducing online harms, and lists commitments to prevent obviously unlawful or high-risk outputs from being generated or displayed, and to detect and act quickly when problematic content appears.
Supporting quotations
We balance access to useful information with the need to reduce online harms.
Prevent obviously unlawful or high-risk outputs from being generated or displayed.
Detect and act quickly when problematic content appears.
Qualifications & exceptions
The prevention commitment covers outputs that are "obviously unlawful or high-risk", not all harmful content.
"Quickly" is not defined.
Not stated here
The policy does not define "high-risk" or set a response time.
Security & Safeguards · Company commitment
Manus commits to be transparent about its safety approach and to provide channels for user feedback
Among its safety commitments Manus lists being transparent about its safety approach and providing channels for user feedback. The reporting section later gives email contacts for concerns.
Supporting quotations
Be transparent about our safety approach and provide channels for user feedback.
Not stated here
The policy does not say how or how often Manus reports on its safety approach.
Security & Safeguards · Company practice
Manus uses automated detection for certain categories of illegal content, including CSEA
Manus says it uses automated detection for certain categories of illegal content, naming CSEA as one. The wording limits detection to certain categories rather than all illegal content.
Supporting quotations
Detection mechanisms: We employ automated detection for certain categories of illegal content, including CSEA.
Automated detection for certain categories of illegal content, including CSEA;
implement automated detection for certain illegal content categories.
Qualifications & exceptions
Covers "certain categories" of illegal content, not all illegal or harmful content.
Not stated here
The policy does not list the other categories, say whether detection covers user inputs, outputs or stored files, or say what happens to the data it processes.
Once Manus becomes aware of illegal content, it says it takes action to address it
Manus says that once it becomes aware of illegal content, it takes action to address it. The trigger is Manus becoming aware, whether through detection or a report.
Supporting quotations
Removal: Once we become aware of illegal content, we take action to address it.
Qualifications & exceptions
Triggered when Manus "become[s] aware" of the content.
Not stated here
The policy does not say what action is taken or how fast.
Security & Safeguards · Permission or license
Manus may refuse, redact or transform content that is likely to cause harm
Beyond illegal content, Manus says it seeks a safe experience for all users and may refuse, redact or transform content that is likely to cause harm, giving examples. It is a permission ("may"), and the examples are introduced by "such as".
Supporting quotations
We seek to provide a safe experience for all users. We may refuse, redact, or transform content that is likely to cause harm, such as:
Qualifications & exceptions
"May", not always.
Applies to content "likely to cause harm".
Examples are introduced by "such as", so the list is not stated to be complete.
Not stated here
The policy does not say how likely harm is judged.
Security & Safeguards · Context
Harmful-content examples span sexual content, self-harm, hate, graphic violence, drug misuse and body shaming
The listed examples of harmful content Manus may refuse, redact or transform are: pornography or sexualised content; encouragement or instructions for suicide, self-harm or eating disorders; hate speech, bullying and harassment; realistic graphic violence; promotion of serious violence or dangerous challenges likely to cause injury; drug misuse encouragement; and content that promotes body shaming or depression. These are framed as content Manus may refuse or change, not as a flat ban.
Supporting quotations
Pornography or sexualised content;
Encouragement or instructions for suicide, self-harm, or eating disorders;
Hate speech, bullying, and harassment;
Realistic graphic violence;
Promotion of serious violence or dangerous challenges likely to cause injury;
Drug misuse encouragement;
Content that promotes body shaming or depression.
Qualifications & exceptions
These examples sit under the "may refuse, redact, or transform" statement, not a prohibition list.
Graphic violence is listed as "realistic" graphic violence.
Not stated here
The policy does not say whether sexual content is prohibited or only subject to refusal; the image, audio and video note separately prohibits sexual content involving minors.
Cross-references
contrast (model-proposed): Generating sexual content involving minors, graphic violence and other illegal content is prohibited in media — Realistic graphic violence is listed among content Manus may refuse, redact or transform, while the image, audio and video note says Manus prohibits generation of graphic violence among other illegal categories. The wording differs in strength and scope and the policy does not say how they fit together.
Security & Safeguards · Company commitment
Manus says it clearly identifies itself as an AI agent in user interactions
Manus says it clearly identifies itself as an AI agent when interacting with users.
Supporting quotations
Manus clearly identifies itself as an AI agent in user interactions.
Not stated here
The policy does not say how the identification is shown.
Security & Safeguards · Company practice
Manus uses human review where appropriate for reported content and escalations
Manus lists human review among its safety measures, used where appropriate for reported content and escalations.
Supporting quotations
Human review where appropriate for reported content and escalations.
Qualifications & exceptions
Used "where appropriate", for reported content and escalations.
Not stated here
The policy does not say what counts as an escalation or what reviewers can see.
Security & Safeguards · Procedure
Report illegal or harmful content by emailing support@manus.im, or feedback@manus.im for US users
If you encounter illegal or harmful content in Manus, the policy asks you to contact support@manus.im, or feedback@manus.im for US users, and to include relevant context such as what you requested, what content appeared, and timestamps where available.
Supporting quotations
If you encounter illegal or harmful content in Manus, please contact us at support@manus.im (or feedback@manus.im for US users).
Please include relevant context such as what you requested, what content appeared, and timestamps where available.
UK users may submit complaints regarding illegal content via the contact information in Section 8.
Qualifications & exceptions
UK users may submit complaints regarding illegal content via this contact information (the regional note points to "Section 8").
The feedback address is given for US users in a parenthetical.
Timestamps are asked for "where available".
Not stated here
The parenthetical does not make clear whether US users must use feedback@manus.im instead of, or in addition to, support@manus.im.
The reporting section is the eighth of eleven headings by position and the saved page shows it with the label "8."; the extracted text does not carry the number.
Manus says it will review reports and take appropriate action, which may include removal or account restrictions
Manus says it will review reports and take appropriate action. Possible actions are content removal, account restrictions, or cooperation with authorities as required by law.
Supporting quotations
We will review reports and take appropriate action, which may include content removal, account restrictions, or cooperation with authorities as required by law.
Qualifications & exceptions
Possible actions are introduced by "may include".
Cooperation with authorities is "as required by law".
Not stated here
The policy does not say how long review takes or whether the reporter is told the outcome.