Groq Acceptable Use and Responsible AI Policy

GroqCloud · agreement dated October 15, 2025 · Official agreement

Read the agreement

3 of 17 cards in Security & Safeguards

Security & Safeguards · Restriction or exception

Attacks on systems, malware, phishing and cyber operations are prohibited

The Customer may not use the services to gain unauthorized access to, attack, abuse, interfere with, intercept, disrupt or exploit any users, systems or services. The policy names offensive cyber operations and automated vulnerability discovery and exploitation as examples. It separately bars phishing and generating or distributing malware such as viruses, worms, Trojan horses or spyware, or other items of a destructive or deceptive nature.

Supporting quotations

to gain unauthorized access to, attack, abuse, interfere with, intercept, disrupt, or exploit any users, systems, or services, including any offensive cyber operations or automated vulnerability discovery and exploitation;
for phishing, generating, or distributing malware such as viruses, worms, Trojan horses, or spyware or other items of a destructive or deceptive nature;

Qualifications & exceptions

  • The systems and users listed are not limited to Groq's own.

Security & Safeguards · Restriction or exception

The services may not be used in ways that burden or impair them

The Customer may not use the services in a manner that burdens, disables, impairs or interferes with the Cloud Services or AI Model Services or the equipment used to provide them. The prohibition covers the services and the equipment behind them.

Supporting quotations

in a manner that burdens, disables, impairs, or interferes with the Cloud Services or AI Model Services or the equipment used to provide the Cloud Services or AI Model Services; or

Not stated here

  • The policy does not say how much load counts as burdening the services.

Security & Safeguards · Restriction or exception

Intentionally circumventing protections or prompting models to break the rules is prohibited

The Customer may not use the services to intentionally circumvent any aspect of the Cloud Services or AI Model Services, including abuse protections or safety mechanisms, if any. The Customer also may not use them to prompt models to act in a manner that violates Groq's policies, this AUP or the Groq Services Agreement. The circumvention item applies to intentional acts.

Supporting quotations

to intentionally circumvent any aspect of the Cloud Services or AI Model Services, including abuse protections or safety mechanisms, if any, or to prompt models to act in a manner that violates our policies, this AUP, or the Groq Services Agreement.

Qualifications & exceptions

  • Circumvention must be intentional; the prompting limb has no stated intent requirement.

Not stated here

  • Groq's other policies, which prompting must not violate, are not listed or supplied here.