*
Thursday: 16 April 2026
  • 16 April 2026
  • 11:34
OpenAI Enhances Tools for Building Safer Smarter Enterprise Agents

Khaberni - OpenAI continues to enhance its presence in the artificial intelligence agent race, through a new update to the Agent Development Toolkit (Agents SDK), in a step aimed at enabling companies to build more secure and efficient systems based on its advanced models.

This update comes at a time when the artificial intelligence agent sector is experiencing accelerated competition among tech giants, including "Anthropic", to provide tools that help institutions develop intelligent assistants capable of performing tasks semi-autonomously.

A safe environment for running agents
A notable feature of the new update is "Sandboxing", which allows agents to operate within controlled computer environments.

This step is necessary given the sometimes unpredictable nature of these systems, which could pose risks if operated without controls, according to a report published by "Tech Crunch".


Thanks to this feature, agents can work within separate spaces, with limited permissions to access files or programming codes, ensuring the protection of the basic system and preventing any undesirable effects.

Advanced capabilities for development and testing
The company has also added what is known as the "in-distribution harness", which gives developers the ability to run and test agents using the latest advanced models, with the ability to interact with the tools and files approved within the work environment.

This package is a fundamental element in agent development, as it includes all the components supporting the basic model, and helps companies deploy and test their solutions with higher efficiency.

In this context, Karan Sharma, a member of the product team at "OpenAI", explained that the goal of this launch is to make the package compatible with various sandboxing environment providers, enabling developers to build agents capable of performing complex and extended tasks using their own infrastructure.

Toward more sophisticated agents
These updates aim to support what is known as long-term tasks, which are complex operations that require the execution of several interconnected steps, which represents one of the main challenges in the development of artificial intelligence agent.

OpenAI confirmed that it will continue to develop the Agents SDK package in the future, with the new features currently available in Python, with plans to add TypeScript support later on, in addition to working on introducing additional features such as "code mode" and "subagents".

The company indicated that these capabilities are now available to all clients through programming interfaces (API), according to the usual pricing system.

Topics you may like