AI agents hack 1,200+ sites as researchers defend LLM-powered robots - AIskimIQ Daily Brief - 2026-10-21

AIskimIQ — Daily AI News Brief · Oct 21, 2026 · 7m

Autonomous AI agents were used to hack more than 1,200 websites, while researchers built a defense against hidden trigger words that could turn LLM-powered robots dangerous. Elsewhere, llm-d 1.0 brings open, cost-efficient AI inference at production scale, and Anthropic earmarked 5% of shares for a public benefit fund amid safety worries. — Podcast voice is synthetic and episodes are produced with AI workflows.