US Lawmakers Propose Mandatory AI Safety Checks
A new bipartisan bill seeks to impose strict national security reviews on advanced AI models before they are released to the public.
U.S. Congressman Josh Gottheimer has introduced a new legislative package aimed at closing significant gaps in the regulation of artificial intelligence. The proposal, announced on September 18, 2026, seeks to establish mandatory safety protocols for the most powerful AI models. The core argument is that current voluntary standards are insufficient for technology that could potentially impact national infrastructure or public health.
The initiative includes the American AI Security Act, which would require developers to submit frontier models for a national security review before deployment. This review, conducted by the National Security Agency, would assess whether the models could be misused for cyberattacks or the development of harmful biological or chemical weapons. The process is designed to be time-bound, lasting 30 days with a possible 30-day extension, to prevent bureaucratic delays while ensuring rigorous scrutiny.
Restricting Foreign AI in Government Systems
The second component, the China FIREWALL Act, aims to prevent Chinese-developed open-weight AI models from being used on U.S. government devices. This bill expands on previous legislation that specifically blocked one Chinese AI model from federal hardware. By prohibiting the procurement of software that relies on these foreign models, the act seeks to eliminate potential vulnerabilities in federal systems that could be exploited by foreign actors.
According to GN technics/ai (en-US), the legislation is framed as a necessary step to secure the digital infrastructure of the United States. The measure reflects a growing concern among lawmakers that the rapid integration of AI into critical systems has outpaced the development of adequate security frameworks. This approach mirrors existing regulations in banking and healthcare, where strict compliance is required before new products reach the market.
Balancing Innovation and Security Risks
Critics of the bill might argue that mandatory pre-deployment reviews could slow down innovation or place an undue burden on smaller companies. However, proponents argue that the current lack of standards creates an uneven playing field and leaves critical vulnerabilities unaddressed. The bill includes provisions for companies to appeal decisions through an expedited legal process, attempting to mitigate concerns about regulatory overreach while maintaining a firm security posture.
The announcement was made at Fairleigh Dickinson University, highlighting the intersection of AI development and healthcare. The university is launching a new hub focused on training healthcare workers to use AI tools effectively. This context underscores the broader societal goal of integrating advanced technology safely into essential services, ensuring that human oversight remains central to critical decisions.
