OpenAI has launched an updated version of its Agents SDK that introduces new features for building enterprise-grade AI agents. The toolkit now includes enhanced safety checks and improved capability controls to address growing concerns about reliability in business applications.
The update focuses on making agents safer for commercial use. It adds automated risk assessments that flag potential issues before deployment. Developers can now set stricter guardrails for agent behavior, reducing the chance of unexpected actions in production environments.
The new version also expands the API to support more complex workflows. Enterprises can integrate agents with existing systems more easily, improving efficiency in tasks like customer support and data processing. OpenAI says early users report faster development cycles and fewer errors in testing.
This follows last month’s announcement of GPT-5, which introduced native agent capabilities. The company now positions the SDK as a key tool for businesses adopting AI agents at scale. Competitors like Google and Anthropic are also releasing similar frameworks, signaling a shift toward enterprise-focused AI development.
The update arrives as companies face stricter regulations on AI safety. OpenAI highlights the SDK’s compliance tools, which align with emerging standards in Europe and the U.S. The company plans further refinements based on user feedback.
Source: techcrunch.com