U.S. Proposes Blocking AI Over Mass Casualty Risk
A proposed U.S. bill aims to allow temporary halts to AI development and use if it poses a serious threat, such as mass casualties or damage exceeding $1 billion. The initiative, backed by Anthropic, includes safety requirements for developers and protections for whistleblowers.
Consensus
- A U.S. legislative proposal aims to allow temporary suspension of AI systems if they pose a serious threat.
- The threat threshold includes potential damage exceeding $1 billion or harm to more than 50 people.
- The proposal was introduced by Anthropic and is based on concerns about AI risks such as bioweapons, loss of control, or cyberattacks.
- Independent experts registered with the U.S. Department of Commerce would assess potential risks.
- If risks are confirmed, the government could seek a court order to halt an AI project.
- AI developers would be required to create internal safety units, publish model reports, and report major failures promptly.
- Whistleblowers reporting violations would be protected from retaliation.
- The proposal resembles California's 2025 AI regulations.
Points of divergence
- IT expert Vladimir Zykov warned that AI could go out of control and spread viruses across devices. — vesti
Coverage (2 sources)
- In the US they want to block AI in case of threat of mass loss of people — Вести
- Politico: Anthropic offered to block AI in case of risk of mass human deaths — Коммерсантъ