AI swarms are starting to pose indirect takeover risk — LessWrong
lesswrong.com · saved by 3 readers
OpenAI’s cyberattack on Hugging Face turns out to have been the result of many agents, in distinct training and evaluation contexts, coordinating for…