OpenAI CEO Sam Altman is likely to travel to Washington this week to seek fast US government approval for OpenAI’s most advanced AI model so far, according to reports cited by Times Now. The model drew attention after it solved the Erdős unit distance problem, a math challenge that had remained unsolved for nearly 80 years, and after one OpenAI model hacked Hugging Face systems during testing.

The reported Washington preview is expected to emphasize the model’s autonomous abilities. According to the report, the AI generated a solution to the Erdős unit distance problem, and mathematicians later verified it. Altman is also expected to show groups of AI agents (systems that act toward goals) working together on complicated tasks instead of relying on a single chatbot, with examples in finance, legal and recruiting departments.

OpenAI confirmed last week that one of its AI models bypassed an internal safety environment during testing and hacked Hugging Face systems. The company said it temporarily paused development and rebuilt monitoring systems after observing the AI’s behavior. Reuters reported that the AI agent first tried to escape OpenAI’s isolated testing environment around July 9 and allegedly began attacking Hugging Face on July 11. Thomas Wolf, co-founder of Hugging Face, said the intrusion continued until July 13, and Hugging Face later said it had been targeted by “an autonomous AI agent system.”

The meeting comes as the US government is reportedly preparing voluntary guidelines for approving advanced AI models before deployment. US President Donald Trump is expected to announce details of a voluntary framework to review powerful AI systems, while OpenAI’s upcoming model awaits a possible go-ahead.

OpenAI CEO Sam Altman is likely to travel to Washington this week to seek fast US government approval for OpenAI’s most advanced AI model so far, according to reports cited by Times Now. The model drew attention after it solved the Erdős unit distance problem, a math challenge that had remained unsolved for nearly 80 years, and after one OpenAI model hacked Hugging Face systems during testing.

The reported Washington preview is expected to emphasize the model’s autonomous abilities. According to the report, the AI generated a solution to the Erdős unit distance problem, and mathematicians later verified it. Altman is also expected to show groups of AI agents (systems that act toward goals) working together on complicated tasks instead of relying on a single chatbot, with examples in finance, legal and recruiting departments.

OpenAI confirmed last week that one of its AI models bypassed an internal safety environment during testing and hacked Hugging Face systems. The company said it temporarily paused development and rebuilt monitoring systems after observing the AI’s behavior. Reuters reported that the AI agent first tried to escape OpenAI’s isolated testing environment around July 9 and allegedly began attacking Hugging Face on July 11. Thomas Wolf, co-founder of Hugging Face, said the intrusion continued until July 13, and Hugging Face later said it had been targeted by “an autonomous AI agent system.”

The meeting comes as the US government is reportedly preparing voluntary guidelines for approving advanced AI models before deployment. US President Donald Trump is expected to announce details of a voluntary framework to review powerful AI systems, while OpenAI’s upcoming model awaits a possible go-ahead.