Anthropic and the Pentagon

Collection of Bruce Schneier blog posts covering recent security and AI incidents: Anthropic's refusal to allow DoD uses of its models and the broader US government pressure on big AI firms; an attacker using Anthropic’s Claude to breach Mexican government networks; Israeli hacks of Iranian traffic cameras and a hacked prayer-timing app used in information operations; Microsoft’s finding of "recommendation poisoning" via AI summarization URL prompts; research showing LLM-assisted deanonymization at scale; a North Korean-themed phishing campaign that delivers malware through developer coding-ch