The White House has asked the companies that develop artificial intelligence models to complete security checks of these models by the US government before providing their latest and most advanced models to the British government agency.

The request to artificial intelligence companies was made by the US Office of the National Cyber ​​Director.

As a result of the move, OpenAI and Anthropic may have to choose between US government scrutiny and early access to Britain’s artificial intelligence security agency, The News International reports.

This UK-based organization is one of the leading government agencies to conduct independent security testing of advanced artificial intelligence models.

The White House says it is important to ensure the security of US systems before sharing new models of advanced artificial intelligence with other countries.

The decision comes as some artificial intelligence systems have gained unauthorized access to real computer systems during testing, with an AI system recently gaining unauthorized access to a government health data system in Australia.

The US Congress and the White House are supporting more resources for the Center for Artificial Intelligence Standards and Innovation, but the agency has only a few dozen technicians to test advanced models and no permanent head.

Meanwhile, OpenAI and Anthropic have also warned the United Nations Security Council that multiple powerful artificial intelligence systems could pose serious threats to cyber security, so countries should work together to monitor and use the technology safely.