Scope and review
AI-generated outputs can contain errors, omissions, or inappropriate content. Review important outputs before relying on them. These proposed terms do not guarantee accuracy or replace professional judgment.
Customer instructions and decisions
Customers should provide lawful instructions, assess output fitness, and decide whether an automated action is appropriate. Assign human oversight according to the consequences of the task, particularly for high-impact decisions.
Agents, permissions and tools
Limit agents to authorized data and tools. Validate workflows before enabling actions that affect people, transactions, permissions, or records. Define approval points and a human handoff. This draft does not claim that any particular control is implemented in every product.
Prohibited uses
Do not use AI to evade permissions, unlawfully discriminate, deceive recipients, compromise systems, or perform unauthorized acts. The Acceptable Use Policy also applies when incorporated into the final agreement.
Providers and data handling
Model providers, information transmitted to them, retention, training or improvement uses, and any opt-out settings require verification. This draft makes no claim that customer data is or is not used for model training. Confirm those practices before providing confidential or sensitive information.
Responsibility and support
The final agreement must address allocation of responsibilities, output rights, supported use cases, and escalation. A verified contact channel is not yet available in this standalone build. You can prepare a request on the Contact page; preparing it does not send it.