The GAO report is the basis for the next steps, but the full filing and the complete list of test prompts are not public in the sources available. No new legislation has been reported. Sen. Maggie Hassan, D-N.H., who takes over as chair of the Senate Homeland Security Committee in the new year, asked the GAO to study how commercially available AI could be turned to malicious ends and what national security threats that poses. Speaking to reporters, she said the danger is no longer theoretical and that "We need to act now." Hassan is preparing hearings on the subject for the new Congress. Anthropic, which makes Claude, told the GAO it does not dispute that a determined tester can extract harmful content by misstating a request, and said it takes the finding seriously. Most of the major developers said their models are still being refined. The GAO noted that providers will keep improving safeguards as capabilities advance, while warning that the technology's benefits could be exploited by bad actors.
Attorneys tracking AI liability, critical infrastructure security or terrorism-related regulation should watch the timing. The shift in committee control sets up Senate oversight of AI safety and terrorism risk in the next Congress, and the hearings will rest on a formal watchdog report rather than press speculation. Watch for the GAO's full report, the hearing schedule, and any bills that would impose safeguard requirements or testing obligations on model developers. Companies facing infrastructure-related claims or compliance questions should review their own red-teaming and misuse-prevention documentation now, before the hearings put it in the spotlight.