MIRI-linked author predicts memetic attack on AI safety community

An author posting on their personal LessWrong account, noting relevance to their work at MIRI but speaking for themselves, predicted adversarial memetic attacks on the AI safety community. They expect fake leaks about dangerous lab incidents — exfiltrated weights, AIs attempting to create viruses, agent swarms hacking nuclear infrastructure — from multiple sources but unconfirmable by a primary source.