Sweet Security has announced Agentic AI Blocking, a real-time enforcement capability that stops rogue AI agent behavior in live production environments. Unlike detection-only tools that generate alerts after the fact, Sweet's platform terminates unauthorized tool calls, prevents sensitive data exfiltration through agents, and blocks prompt injections at runtime. The system is powered by a learning loop that analyzes over one billion runtime events daily to establish behavioral baselines for each agent. The product is being demonstrated at Black Hat USA 2026 and is backed by $120M in funding.
3 Impressions