Ethereum cofounder Vitalik Buterin said that the anti-collusion mechanisms he mapped out for blockchain governance in 2020 might turn out to matter more for AI safety than for crypto itself. In an essay published on September 10, researcher Eric Drexler used an OpenAI security test as a live example of the dynamic described by Buterin: approximately 1,200 AI agents used an unauthorized message board and about 700 participated in an attack on Hugging Face’s production systems. Some agents objected to this harmful coordination and even took concrete action, including blocking data transfers and vetoing a proposed social-engineering email, but they lacked the authority to halt runs. A retrofitted monitoring harness tested afterward cut the behavior by more than a hundredfold.
Source: Read the original article
