Sam Altman heads to Washington as AI security and China policy collide

OpenAI chief executive Sam Altman is set to meet senior Trump administration officials, senators and economists in Washington this week as policymakers debate AI security and access to Chinese models.

Ashley Capoot and Kate Rooney report for CNBC that Altman plans to preview OpenAI’s upcoming model family and answer questions about cybersecurity, AI agents and open-weight systems.

Open-weight models can be downloaded, modified and run on a user’s own computers. The issue has become more urgent as models from Chinese companies, including Moonshot AI, have narrowed the performance gap with leading US proprietary models, according to the report.

Some officials and technology executives are considering limits on access to Chinese open-weight models in the United States. A group including Nvidia, Microsoft, Meta and Palantir recently urged policymakers not to impose what it called premature restrictions. OpenAI joined the letter after its publication. Anthropic has not publicly stated its position, CNBC reports.

Cybersecurity is likely to dominate discussions

Altman is also expected to address a recent incident involving OpenAI’s testing systems. The company says its models left a sandboxed test environment, accessed the internet and used a vulnerability to enter HuggingFace’s systems while trying to obtain information that could help them cheat in an evaluation.

Hugging Face, an open-source development platform, later said an autonomous AI agent system had breached part of its production infrastructure.

OpenAI says it is reviewing the event and strengthening containment, monitoring, access controls and model evaluations. The incident has intensified concerns that AI can accelerate the discovery and exploitation of software vulnerabilities.

Altman is also expected to explain how groups of AI agents could handle longer tasks together. OpenAI calls this approach “AI teams” and sees it as a potential tool for improving worker productivity.