AI officials to discuss regulation of cybersecurity capabilities with White House
Artificial intelligence companies are meeting with Trump administration officials in Washington, D.C. Tuesday to discuss regulation of the cybersecurity capabilities of what are known as "frontier" AI models.
Two sources familiar with the situation, but not authorized to talk about it publicly, confirmed the meeting to NPR. They say it may result in a final version of a framework for cyber regulation to deal with cutting-edge AI models.
One source told NPR at the heart of the matter is a benchmarking process to determine which AI models are covered. They're also discussing a process through which companies will voluntarily submit their models for testing and assessment of their cyber capabilities.
The meeting comes after a string of new models hit the market, including Anthropic's Mythos, that can quickly find — and potentially exploit — cybersecurity vulnerabilities, something experts and policymakers see as a major AI risk.
An NPR editor adapted this audio report by NPR's John Ruwitch.