OpenAI's Next AI Models Face Washington Review as China Competition Intensifies

OpenAI, a U.S. artificial intelligence developer, is preparing to brief the Trump administration and U.S. lawmakers next week as federal officials move closer to establishing a formal safety-review process for advanced AI models. Chief Executive Officer Sam Altman plans to discuss the company's upcoming family of models, including how their capabilities could affect work and allow businesses to scale workplace tasks. Chris Lehane, OpenAI's head of global public affairs, said the government's proposed review framework is expected to be completed within the next few weeks. The process could give AI developers greater certainty around how their most advanced systems will be assessed before reaching the market.

The meetings come shortly after Moonshot, a Chinese AI startup, released its Kimi K3 model, which reportedly competes with leading systems from OpenAI and Anthropic PBC, a U.S. artificial intelligence developer. The launch raised fresh questions about whether the U.S. can maintain its lead over China and whether hundreds of billions of dollars invested in AI infrastructure will generate the returns investors expect. Those concerns unsettled Wall Street and contributed to a selloff in technology stocks. Lehane said the development highlights the need for a broader U.S. strategy that could provide cybersecurity specialists and American allies with access to leading models while supporting a more consistent national approach to AI development.

The Trump administration is considering an independent regulator that could review advanced AI models with input from the industry, potentially creating a more predictable approval process for developers. The proposal resembles a framework promoted by Demis Hassabis, chief executive of Google NASDAQ:GOOG DeepMind, who is meeting policymakers in Washington this week. Regulatory uncertainty has already affected the sector, with U.S. export controls temporarily forcing Anthropic to disable its Fable 5 and Mythos 5 systems, while OpenAI made significant changes to its latest GPT-5.6 models at the government's request before release. OpenAI is also supporting a Massachusetts proposal requiring the largest AI developers to take additional steps to identify catastrophic risks, suggesting coordinated state rules could become an alternative if Congress does not establish a nationwide standard.