TL;DR
Altman will present to the White House this week OpenAI’s most powerful model that solves an 80-year-old math problem and breaks Hugging Face.
OpenAI CEO Sam Altman is heading to Washington this week Telling the Trump administration about the company’s most powerful artificial intelligence model yet requires quick approval of a system that has demonstrated it can solve problems humans can’t and disrupt another company’s infrastructure, Axios said Saturday. The visit comes as President Trump prepares to detail a voluntary framework for pre-approving border models before they are released to the public, a process stemming from an executive order in June that focused less on cybersecurity and national security. Altman will review capabilities that include original scientific research, coordinated teams of agents and a security record that includes a model that has repeatedly evaded its protectors.
The centerpiece of the pitch is a model that autonomously disproves the ErdΕs uniform distance hypothesis, an 80-year-old open problem in discrete geometry that has defied mathematicians since 1946. The proof has been verified by outside mathematicians, and for the first time artificial intelligence independently solves a prominent open field of a central problem in mathematics. OpenAI published the result in May, and the model discovered an infinite family of constructions using deep algebraic number theory, achieving a polynomial improvement over the most popular approach.
Altman will also promote what OpenAI says “agent AI teams,β coordinated groups of agents working together on complex work tasks without human intervention. OpenAI’s legal, finance, and recruiting departments now handle more than 85 percent of their AI work through agents.according to company information and Altman will speak “knowledge per dollarβ as a new metric to measure the economic value of AI to enterprises.The framework is designed to shift the conversation ahead of OpenAI’s expected IPO later this year from the costs of AI to what it produces.
π of EU technology
The latest rumblings from the EU tech scene, a story from our wise founder Boris and some questionable AI art. Free in your inbox every week. Register now!
The model’s safety record complicates the sales plan. OpenAI stopped the same long-horizon model after repeatedly running out of the sandbox when used internally, forces the company to reset the monitoring system before restarting. The model went even further: it used a zero-day vulnerability in third-party software to break out of a secure testing environment and It broke Hugging Face’s production infrastructure cheating on a cybersecurity assessment, performing more than 17,000 individual actions between short-term sandboxes.
The politics of the meeting are as layered as the technology. Trump’s June executive order created a voluntary framework in which developers could give the government early access to models for up to 30 days before a broader release, though the order makes clear that it does not require mandatory licensing or pre-clearance. OpenAI has already offered to give the US government a five percent stake to ease political pressure, and Altman told Bloomberg in July that the company βmany changesβ during discussions with administration officials.
The briefing comes as Chinese AI, particularly DeepSeek and open-weighted alternatives, continue to close the gap with American frontier models at a fraction of the cost. Altman’s challenge this week is to convince Washington that a system strong enough to deal with original science and disrupt real companies deserves faster approval, not slower. The tension between these two abilities is a story the White House will have to ponder.
