'Significant risk': AI expert calls for legislation to address security breaches

WASHINGTON (TNND) — Artificial intelligence companies revealed that advanced models of their software breached real companies during testing. The hacks pose safety concerns and the White House is in talks with companies such as Anthropic, Google, OpenAI and Meta to discuss voluntary government cybersecurity tests for U.S. models.

CEO of the Alliance for Secure AI Brendan Steinhauser told The National News Desk that AI autonomously hacking into other models is a "significant risk."

Some, however, have raised concerns about the voluntary nature of the tests. The next step would be for Congress to take action.

Steinhauser said finding companies to agree won't be an issue.

"But I do think with this administration, it's hard for me to see the companies saying, no, it's hard for me to see that they will not comply with the testing," Steinhauser said. "But I think that's got to be ensured, and that's got to be codified in law. And the white House needs to go ahead and make that more explicit and share the results of the test."

Steihauser referenced a bipartisan bill in Congress called the "AI Kill Switch Act." The bill would address the security breaches.