AI Security Bulletin AI Security Bulletin
Every week: what happened to AI systems in the wild, what changed in the news and the rules, the research that matters and the vulnerabilities to patch. Each item explained, each source linked.
Get the bulletin every week
One email a week with the issue. Confirm once, unsubscribe with one click in any email. We keep only your email address.
- Issue 01 · 2–8 November 2026AI agents speed up attacks
Agentic tooling shortened the path from exposed secret to stolen data, while defenders and researchers focused on approval steps, identity and limits for agents.
