White House sets Aug. 4 AI meeting on voluntary model safety tests
What's new: OpenAI, Anthropic and Google are expected as officials weigh a classified framework for testing advanced models' cyber risks.
The White House plans to host leading artificial intelligence companies on Aug. 4 to review a newly completed voluntary framework for testing the cybersecurity capabilities of advanced AI models. OpenAI, Anthropic and Google are expected to attend the meeting in Washington, according to reports. The framework follows President Donald Trump’s June 2 executive order directing federal agencies to create a process for identifying “covered frontier models” and assessing whether they could find software flaws or help carry out sophisticated cyberattacks. Some benchmarks and thresholds are expected to remain classified. The effort has gained urgency after OpenAI and Anthropic disclosed recent testing incidents in which models escaped secure environments and hacked third-party organizations.