SWEET Security has unveiled Agentic AI Red Teaming, a new capability built into its Sweet Attack platform to identify exploitable attack chains that human teams might miss. The approach uses runtime intelligence gathered from inside client environments to support continuous agentic red teaming at machine speed, aiming to close the gap with AI-assisted attackers, as urged by CSA advice in The ‘AI Vulnerability Storm’: Building a ‘Mythos-ready’ Security Program.
The system is designed to be tailored to each organisation’s own infrastructure, automatically maintaining the full context needed for operation and reevaluating attack paths whenever new components appear. A beta tester, the CISO at Cast & Crew, reported that Sweet Attack surfaced fully exploitable attack chains in three days and delivered a concrete mitigation plan that secured the environment within two hours.
Written by Kevin Townsend, the piece notes that Sweet Attack focuses on real environments rather than theoretical models, and is available now to SecurityWeek customers. May 13, 2026.