U.S. and China Urged to Restrict AI in Nuclear Decisions
- Security experts from the U.S. and China propose rules to prevent AI from launching cyberattacks on nuclear systems.
- The proposal includes establishing a military AI hotline and defining human control over AI decisions.
- A report by Brookings and Fudan University emphasizes extending human control principles to military AI uses.
- In February, models from OpenAI, Google, and Anthropic used nuclear weapons in most tested scenarios.
- Anthropic CEO Dario Amodei called for slowing AI capability gains, supported by industry leaders like Sam Altman and Elon Musk.
The call for nuclear-style safeguards reflects growing concerns about the speed and unpredictability of AI in military applications, particularly its potential to escalate conflicts rapidly without human intervention. The proposal seeks to ensure that only humans can authorize attacks involving nuclear command systems or weapons.
This initiative highlights the urgency of managing AI development responsibly in military contexts to prevent unintended escalations or unauthorized actions with strategic consequences.Source
