Sam Altman will travel to Washington next week to brief Trump administration officials and lawmakers on OpenAI’s upcoming models, as the government moves toward a formal framework for reviewing frontier AI systems before they reach the public.

What’s Happening in Washington

Chris Lehane, OpenAI’s global affairs chief, said the discussions will cover the new model family’s capabilities and how they could affect work, telling reporters in Washington that the company expects “really interesting capabilities” particularly around scaling work-related tasks, according to Bloomberg. The trip follows a pattern: Altman told CNBC earlier this month that OpenAI made “many changes” to its models after discussions with officials including Commerce Secretary Howard Lutnick and Treasury Secretary Scott Bessent ahead of GPT-5.6’s public release.

The Proposed Oversight Framework

The visit lands as the administration considers creating a standalone body to evaluate AI systems before launch. One version of that proposal, developed with Bessent’s input, would give AI companies clearer expectations for how their models get assessed a shift from the voluntary, case by case testing arrangements the government has relied on so far. Google DeepMind CEO Demis Hassabis is pursuing a parallel track, meeting Washington offices this week to advocate for a similar review framework, Bloomberg reported.

Lehane also addressed what happens if federal legislation doesn’t materialize. If Congress can’t set a national standard, he said, OpenAI’s fallback is a “reverse federalism” approach in which states adopt matching rules rather than each writing its own. OpenAI is already backing a Massachusetts proposal that would require large AI companies to address severe AI related risks a bill Anthropic has also supported, suggesting at least partial common ground between the two labs on state-level policy even as they compete on models.

The push for a coordinated national framework carries a competitive undertone. Lehane pointed to Moonshot AI’s release of Kimi K3, a model he said challenged systems from OpenAI and Anthropic, as evidence that China’s state backed AI strategy is narrowing the gap with U.S. labs and raising doubts about whether American infrastructure spending will pay off as expected. He argued the U.S. needs a unified strategy to preserve its lead and keep advanced models shareable with allies.

Why the Anthropic Dispute Matters Here

That argument carries more weight after last month’s dispute between Anthropic and the government. Authorities ordered Anthropic to block foreign nationals from accessing its Fable 5 and Mythos 5 models, citing national security authorities and, Anthropic later said, a possible jailbreak technique. The company disabled both models for all customers to comply, then restored access on July 1 after the government lifted the export restriction. The episode showed how unpredictable government intervention in AI can be even for companies already cooperating with federal testing programs and gives OpenAI’s push for a defined, predictable review process an obvious selling point heading into next week’s meetings.