JetStream adds pre-execution agent checks

JetStream Security announced Clearance on September 2, a reasoning engine at its AI gateway that evaluates each agent action against an approved task design before letting it execute, rather than logging it afterward. The company’s example: an agent authorized to query a customer record, prepare an attachment, and send an email is each individually legitimate, but adding a blind-carbon-copy to the send is a step Clearance flags as correlated with data exfiltration and denies before it runs. Clearance enters general availability this fall.