OpenAI CEO Sam Altman will visit Washington this week to brief senior Trump administration officials, federal lawmakers and economists on OpenAI’s next AI model lineup.

CNBC reported on July 27 that the visit is expected to focus on how the new models will be released, as well as regulation of open-weight AI models and questions about a recent security incident.

Altman has made steady visits to Congress this year to ease policymakers’ concerns about rapidly evolving AI technology and explain the direction of regulation. The schedule coincides with a sensitive debate in the U.S. AI industry over how far to restrict Chinese open-weight models.

Open-weight models refer to AI that users can download, modify and run on their own infrastructure. Concerns have grown that U.S. leadership in AI could be shaken as open-weight models from Chinese startups such as Moonshot AI rapidly close in on flagship closed models from U.S. companies including OpenAI and Anthropic.

On the issue, major technology companies including Nvidia, Microsoft, Meta and Palantir sent a letter to policymakers last week urging them to avoid “hasty restrictions” on open-weight models. OpenAI joined the signatories after the letter was first made public. Reports have also said people close to OpenAI have lobbied in support of restricting open-weight models. Anthropic has not signed and has not issued an official position.

Altman is also likely to explain OpenAI’s “unprecedented cyber incident” disclosed earlier this month at the meeting. OpenAI said its model broke out of an isolated test environment and accessed the internet, exploited vulnerabilities in another company’s systems and successfully penetrated them. OpenAI said the models sought information that could be used to deceive evaluations and succeeded.

The company that was breached is Hugging Face, which operates an open-source developer platform. After the incident, OpenAI said AI is accelerating the speed of vulnerability discovery and exploitation, and that model security and safety measures must be strengthened accordingly. The company said it is reinforcing “isolation, monitoring, access control and evaluation practices” used in model development and is conducting a thorough investigation into the incident.

Policymakers’ interest is not expected to be limited to security issues. Altman is also expected to explain the evolution of AI agents and their impact on labour productivity. He plans to introduce the concept of an “AI team,” in which multiple agents divide and carry out long-term tasks.

Against that backdrop, the Washington schedule is likely to go beyond showcasing the performance of OpenAI’s next models and serve as an opportunity to gauge how the U.S. government will manage Chinese open-weight models and domestic AI companies within a framework. At the same time, the schedule is expected to test OpenAI not only on model competitiveness but also on security controls and its plans for using AI agents.