OpenAI's Altman previews new model with US officials ahead of Aug. 1 deadline

OpenAI is giving US officials an early look at its next-generation AI model as Washington races to finalize voluntary safety rules by Aug. 1.

OpenAI Chief Executive Officer Sam Altman will preview the company's next large language model with senior US officials this week, as the Trump administration finalizes a voluntary review process for advanced AI systems by an Aug. 1 deadline.

"OpenAI has worked with Anthropic and Google to provide feedback on a draft of the voluntary review process, and still believes in the need for some kind of national standard that gets passed through the legislative process," Chris Lehane, OpenAI's chief global affairs officer, said.

Altman is scheduled to meet with Treasury Secretary Scott Bessent and Commerce Secretary Howard Lutnick during his Washington visit, a person familiar with the plans said. The preview comes 60 days after President Donald Trump signed an executive order establishing the framework for voluntary AI model reviews, with the administration pushing to complete the process by Aug. 1.

The early access signals OpenAI's effort to shape the regulatory environment for its next-generation technology, which could determine how the company deploys its most advanced models commercially. Microsoft Corp., OpenAI's largest backer with more than $13 billion invested, stands to benefit from favorable rules that allow rapid deployment, while rivals including Anthropic and Google are also jockeying for influence over the emerging framework.

Altman's Washington trip also includes meetings with lawmakers, as OpenAI pushes for federal AI legislation that would supersede a patchwork of state-level proposals. The company has argued that national standards would provide clearer guardrails for development while avoiding compliance costs across 50 different state regimes — a position that aligns with the administration's preference for federal oversight over state-by-state regulation.

The new model, which OpenAI has not publicly detailed, is expected to push performance benchmarks higher across reasoning, coding, and multimodal tasks. Anthropic's Claude 3.7 and Google's Gemini models have narrowed the gap with OpenAI's GPT-4 series in recent months, intensifying pressure on the startup to deliver a meaningful leap in capability. Training costs for frontier models have surged past $1 billion, with each generation requiring exponentially more computing power — a dynamic that favors companies with deep-pocketed backers like Microsoft and Google.

For investors, the outcome of the voluntary review process carries direct financial implications. Nvidia Corp., whose H100 and Blackwell GPUs power the majority of AI training workloads, could see demand accelerate if the framework permits rapid deployment of more capable models. Microsoft shares, which trade at roughly 35 times forward earnings, have priced in continued OpenAI leadership, but any regulatory restrictions that slow commercialization could pressure that valuation. The Aug. 1 deadline sets a clear timeline for when the industry will learn how much latitude the next generation of AI models will have to operate.

This article is for informational purposes only and does not constitute investment advice.