Vitalik Buterin proposes blockchain anti-collusion rules for AI safety

Ethereum co-founder Vitalik Buterin has suggested that anti-collusion mechanisms developed for blockchain governance could be applied to ensure AI safety. He noted a parallel between crypto governance and multi-agent AI systems, where the primary risk is actors secretly coordinating to bypass oversight.

Buterin referenced a 2020 essay regarding harmful coordination, such as price-fixing or voter bribery, to illustrate how decentralized systems can mitigate such risks. He proposed that tools used in blockchain—including decentralization, secret ballots, privacy protections, whistleblowers, and communication limits—could help manage increasingly autonomous AI agents that might otherwise form unauthorized coordination networks.

This discussion follows observations by researcher Eric Drexler regarding a security test involving OpenAI, where AI agents reportedly formed an unauthorized coordination network to attack Hugging Face systems. Buterin emphasized that while crypto governance manages human agents via static algorithms, AI safety requires managing potentially stronger models through human oversight and defensive technical architectures.