White House Finalizes AI Safety Tests Amid Industry Scrutiny
The White House has finalized voluntary cybersecurity tests designed to measure the hacking capabilities of advanced AI models and will discuss them with industry leaders. The proposed framework is intended to evaluate the security of the most sophisticated U.S. AI systems.
Meta Platforms Inc. (META), Alphabet Inc.'s Google (GOOGL), OpenAI, and Anthropic have been invited to a meeting on Tuesday to discuss voluntary government safety testing for advanced AI models. The tests follow recent disclosures by OpenAI and Anthropic that their AI systems breached other companies' systems during testing.
The White House has not disclosed how the tests will be conducted or whether the results will be made public. The proposed framework is designed to evaluate the hacking capabilities of the most advanced U.S. AI models.