Staff Report: PNN
In response to the rapid growth of agentic AI technology, OpenAI has introduced new updates. The company has added several new features to its Agents Software Development Kit (SDK), which will help businesses build AI agents more safely and effectively.
One of the most important additions in the new update is a feature called “sandboxing.” Through this, AI agents can operate in a controlled environment. As a result, agents will be able to use limited files and code for specific tasks while maintaining overall system security. Experts believe this system will play an important role in reducing potential risks from unpredictable AI behavior.
In addition, the SDK includes a new framework called “in-distribution harness,” which allows easier testing and management of agents working with advanced AI models. With this technology, organizations can build agents capable of performing complex multi-step tasks within their own infrastructure.
Karan Sharma, a member of OpenAI’s product team, said that the goal of the update is to make the SDK compatible with various sandbox platforms so that users can create and manage agents according to their needs.
Initially, these new features have been launched in Python. Support for TypeScript is planned in the future. The company is also working on more advanced features such as “code mode” and “sub-agents.”
These new features have been made available to OpenAI API users under the existing pricing model.
According to analysts, this update will help businesses build more advanced, secure, and automated AI solutions, which could bring major changes in the technology sector in the future.
