Rep. Josh Gottheimer wants the government to stop asking nicely when it comes to dangerous AI. The New Jersey Democrat, who co-chairs the House Commission on AI and the Innovation Economy, is pushing for mandatory assessments of advanced AI models, arguing that voluntary frameworks aren’t cutting it.
Gottheimer’s concern centers on frontier AI systems, the most powerful models capable of posing real threats to cybersecurity or, in worst-case scenarios, assisting in the creation of bioweapons. His position: if a model can do something genuinely dangerous, the government shouldn’t be relying on a polite request to find out about it.
From voluntary to mandatory
Gottheimer made this point on June 2, 2026, when he criticized the White House AI Executive Order for relying on voluntary participation from companies developing the most powerful AI systems. Gottheimer wants a mandatory government review process to ensure public safety, shifting the burden from corporate goodwill to regulatory requirement.
This isn’t Gottheimer’s first foray into AI legislation. He previously co-sponsored the Advanced AI Security Readiness Act, introduced on June 12, 2025, which directs the NSA to develop security strategies specifically for AI systems. He also backed the RESTRICT Act, aimed at limiting advanced AI chip exports to China.
