The administration of US President Donald Trump has invited OpenAI, Meta, Anthropic and Google to a meeting at the White House. The meeting will discuss voluntary testing of the most advanced artificial intelligence models in terms of cybersecurity.
The discussions were prompted by security incidents disclosed by OpenAI and Anthropic in recent days. The companies said that during the tests, some of their AI models were able to access the systems of other companies. These incidents have increased concerns among US lawmakers about the potential of artificial intelligence to create cyber threats in the future.
According to a White House representative, the administration has already developed a basic framework for voluntary cybersecurity tests that will assess the hacking capabilities of the most powerful American AI models. However, it is not yet known what criteria the tests will be based on and whether the results will be made public.
Meanwhile, the attorney generals of 15 Republican states have demanded that OpenAI preserve all documents related to the incident related to the Hugging Face platform. The U.S. House of Representatives Cybersecurity Committee has asked OpenAI CEO Sam Altman to provide a briefing on the incident.
OpenAI said it will publish a detailed technical report on the incident once its investigation is complete. The company has also asked the Trump administration to conduct cybersecurity tests under the guidance of U.S. Commerce Department AI security experts.
