OpenAI CEO Sam Altman has briefed U.S. government officials on a new AI system that coordinates multiple "agents" to collaborate on complex tasks, with the company expected to reveal its ability to solve previously unsolved math problems by Friday. The multi-agent system, referred to in private meetings as WAPo, enables software programs to operate simultaneously in the background to divide workloads and perform specialized duties usually handled by different professionals.

The private meetings in Washington with White House and Cabinet leaders took place just before an August 1 deadline for a new federal framework requiring the reporting of major advances in AI capabilities. The briefings were part of a broader effort by OpenAI and rival Anthropic to shape the Trump administration's rules on pre-deployment safety evaluations and national security reviews for the industry's most powerful models.