Over recent months, OpenAI has worked with customers and partners to align safety policies across the services that provide access to its models. The company says these updates aim to make usage requirements clearer and easier for developers and businesses to follow.
As models such as GPT‑6 Astra expand what developers and companies can build, OpenAI is emphasizing consistent safeguards across all services that offer its models — whether accessed directly through OpenAI, hosted by partners, or provided via reseller partners.
Pre-deployment processes
Before deploying a model, OpenAI evaluates its capabilities and risks, trains it to meet safety requirements, and tests whether its safeguards hold up when people deliberately try to bypass them. The company describes its approach as combining multiple safeguards designed to reinforce one another.
Post-deployment monitoring
After deployment, OpenAI monitors for potential abuse and investigates concerns. To mitigate cyber misuse, models are trained to refuse harmful requests and systems flag suspicious interaction patterns to guide response decisions. Trusted access programs add another layer of protection by setting eligibility requirements and conditions for access to the most capable models.
Responsibilities of partners and providers
Some safeguards are embedded in the model itself, while others depend on the surrounding systems and the people operating them. Both types must be effective regardless of whether users access the models directly or through partners. OpenAI requires partners that provide access to its models to maintain safeguards consistent with OpenAI’s own, investigate potential violations, and take appropriate action when concerns are identified. Coordination on enforcement is intended to keep restrictions effective across different services offering access.
Practical implementation and customer-controlled options
Putting these policies into practice requires accounting for how different services operate and how customers use them. One example is collaboration with customers on Private Safety Processing, which OpenAI cites as a way to develop effective safeguards while giving customers greater choice and control over their sensitive data.
Updated Usage Policies
OpenAI is updating its Usage Policies, with the changes taking effect on November 20, 2026. The updated policies clarify rules for safe model development and sensitive uses of AI. They also require services that provide direct access to OpenAI models — such as model routers or AI gateways — to obtain OpenAI’s review and approval before offering that access.
Communication, investigation, and proportional enforcement
OpenAI says it will continue to explain what is allowed and restricted, and how customers and partners can raise questions or resolve compliance concerns. The company continues to improve how it detects, investigates, and responds to misuse; misuse can occur through end users, compromised accounts, or leaked API keys, and OpenAI works with customers to understand and address those issues.
SafetyIDs and targeted responses
OpenAI is also working with customers to introduce SafetyIDs, identifiers that customers can provide to help OpenAI distinguish between individual end users during an investigation without OpenAI learning those users’ identities. SafetyIDs are intended to enable more precise responses to potential misuse while reducing disruption to other users.
Enforcement principles
OpenAI states that enforcement decisions are based on documented conduct and risk, that rules are applied consistently, and that responses are proportional to the circumstances. When access is restricted, OpenAI will revisit those restrictions if meaningful changes in conduct, commitments, or safeguards address the underlying concerns.
OpenAI frames these measures as intended to help more people benefit from AI: clearer rules and consistent safeguards should make it easier for developers and businesses to build, compete, and choose services that work best for them.
