OpenAI Updates Its Agents SDK With Enterprise Safety Controls

OpenAI released an updated Agents SDK on April 15 with a batch of new safety features targeted at enterprise deployments. TechCrunch covered the release, which adds guardrails around agent permissions, tool access, and audit trails for agents handling sensitive workflows.

The enterprise pitch was straightforward. OpenAI has been losing enterprise deals to Anthropic on safety and auditability specifically. The SDK update tries to close that gap by giving enterprise security teams more control over what agents can do and better visibility into what they did after the fact.

For teams building internal AI agents rather than shipping agents to end customers, the update is relevant. It’s the kind of release that moves the needle on whether a compliance team greenlights an agent rollout or flags it for further review. For everyone else, inside baseball.

Source: TechCrunch, April 15, 2026