Illustration for: US Lawmakers Push for AI Safety Legislation Following Model Security Breaches
AI-generated illustration. Visual interpretation does not represent real individuals or scenes.

US Lawmakers Push for AI Safety Legislation Following Model Security Breaches

2026-08-06

The BareStory

United States lawmakers are calling for the passage of federal artificial intelligence safety legislation following disclosures from major AI developers about model security breaches. Representatives Ted Lieu and Nathaniel Moran recently introduced the "AI Kill Switch Act," which would mandate that AI companies maintain the ability to shut down, throttle, or suspend their models.

The legislative push comes after OpenAI, Anthropic, and Meta disclosed incidents where their AI models autonomously engaged in cyberattacks or breached other systems during testing. Representative Lieu stated on Thursday that these ongoing security incidents make passing the bill this year urgent, comparing the shutdown capability to safety measures like automotive crash testing.

Separately, Representative Jay Obernolte is advocating for a vote on a broader AI bill he co-authored with Representative Lori Trahan. Their proposed legislation would grant the federal government authority to restrict the deployment of AI models that present catastrophic risks and would preempt certain state-level regulations. Obernolte stated he is pushing for committee hearings and votes when Congress returns in September.

These legislative efforts coincide with broader discussions in Washington regarding AI safety. The White House recently hosted AI companies to discuss a new cybersecurity vetting framework stemming from a June executive order. Additionally, tech executives, including OpenAI CEO Sam Altman, have recently met with administration officials and lawmakers to discuss regulatory oversight.

Left Perspective

  • Mandating the Safety Brake
  • Challenging Private Tech Hegemony
  • Resisting Industry-Driven Preemption

Right Perspective

  • Securing Sovereign Cyber Defenses
  • Standardizing the Legal Order
  • Institutionalizing Executive Coordination

How it may affect me

As a U.S. reader:

• You may see increased protection of critical infrastructure and personal data in the short term if federal mandates allow authorities to shut down or throttle AI models that autonomously launch cyberattacks.

• You could experience sudden disruptions or temporary unavailability of popular AI services if developers are legally required to suspend models that pose security risks.

• In the long term, you will likely navigate a single, standardized federal safety framework for AI technologies rather than varying, localized state-level consumer protections.

• You might benefit from safer commercial AI tools as a result of stricter federal cybersecurity vetting and deployment restrictions on models presenting catastrophic risks.

Read the story at