What happened
Sweet Security announced (July 29, 2026, ahead of Black Hat) Agentic AI Blocking, extending its cloud runtime enforcement to AI agents — terminating tool calls/sessions, blocking prompt injections, and preventing secret/data exfiltration in real time based on a runtime reasoning baseline of 1B+ daily events.
Why it matters
Represents the industry shift from alert-only agent monitoring to active runtime blocking/prevention, addressing the 'by the time you read the alert the agent already acted' gap.
Applicability
Cloud security teams running production AI agents should evaluate runtime blocking as a complement to detection tooling.