5 min read
News Brief Google's Nov 16 AI Threat Tracker reports an exposed `exploit_agent` server that diffs patches, writes exploits and fires them with no human in the loop. It also flags a suspected DPRK cluster abusing Claude Code's settings file.