Clearance extends AI gateway security by authorizing each and every agentic transaction and decides whether an agent's next step should run before it executes.
SANTA CLARA, CA / ACCESS Newswire / September 2, 2026 / JetStream Security, the AI Control Plane for enterprise AI, today announced JetStream Clearance, a reasoning engine that evaluates every AI agent action against an approved design and decides whether it should run before it executes.
Clearance examines every request at the JetStream AI Gateway, maps it to the agent or user making the request, the approved design which authorizes the request, the tool(s) being invoked, and what that specific call is about to do. Clearance, as the name suggests, is required for the request to be processed by the AI Gateway. Bad or dangerous sequences of requests will be denied clearance, preventing rogue agents and runaway AI systems from performing unauthorized actions. Clearance performs its reasoning in flight, ahead of execution, rather than in a log afterward.
The announcement comes as AI gateways become a commodity infrastructure. Rippling, Ramp, Databricks, Microsoft, and IBM have each shipped one in the past three months, joining LiteLLM, Portkey, Kong, and JetStream. Anthropic documents the gateway as the recommended control point for enterprise Claude Code deployments. Gateways route traffic, cache responses, proxy MCP calls, and enforce rate limits.
None of that work determines whether an agent action should happen in the first place.
Enterprises that have tens of thousands of employees today are preparing to run hundreds of thousands of AI agents, each acting under an identity the company issued. Historically, Zero Trust settled whether a person or an application should reach a resource. It assumed a human on the other end, moving at human speed, with the application as the unit of trust. Today, agents break all three assumptions. To be effective, the trust boundary must move to the individual action, evaluated in the moment rather than reconstructed afterward.
JetStream Clearance applies that principle where agents act, at the tool calls an agent makes through the Model Context Protocol. That is where the risk is concentrated today.
Detection and authorization are different jobs. Runtime security platforms see what an agent did and respond at machine speed. Clearance decides whether the action runs at all. The two layers are designed to operate together. JetStream is demonstrating Clearance at Fal.Con this week and named in CrowdStrike's new AI Partner Specialization, launched at Fal.Con 2026.
"Every enterprise I talk to has an AI gateway now, and that is exactly the problem," said Raj Rajamani, Co-Founder and CEO of JetStream Security. "A gateway can tell you where a request went. The question our customers keep asking is whether the agent should have sent it at all, and until now nothing in their stack could answer that. Clearance is how we deliver Zero Trust for AI. It's every action authorized, every time."
Clearance is based on JetStream AI Blueprints, the versioned operational contracts that describe how each agentic system is assembled, including which models, MCP servers, tools, datasets, and identities it uses. Policy binds to the identity when access is requested and reaches inside the tool and down to the individual parameter. A team can enable a server's read tool and block its delete tool, so an agent never holds a permission it was never meant to have.
The distinction matters most in sequence. For example, an agent authorized to query a customer record, prepare an attachment, and send email may take each of those steps legitimately. Adding an external blind carbon copy (bcc) to the final send is a step that has high correlation to data exfiltration. Clearance evaluates the sequence rather than the isolated call and stops the send.
"The industry spent a decade learning that a firewall that only reads ports is not a security control," said Venu Vissamsetty, Co-Founder and Chief Architect of JetStream Security. "We are at that same moment with AI. Enterprises are deploying agents faster than they can describe what those agents are permitted to do, and most are finding the gap the hard way. This is the control our customers have been asking us for."
JetStream Clearance enters general availability this fall. JetStream will demonstrate Clearance at Fal.Con, booth #1414, August 31 through September 2, 2026.
About JetStream Security
JetStream Security is the AI Control Plane for enterprise AI. The platform makes every AI agent, model, workflow, and identity in an enterprise visible, attributed, and governed, continuously and at runtime, across the entire AI estate. JetStream is backed by Redpoint Ventures and the CrowdStrike Falcon Fund.
Media Contact:
Mia Balaban
mia@tellny.com
SOURCE: JetStream
View the original press release on ACCESS Newswire:
https://www.accessnewswire.com/newsroom/en/computers-technology-and-internet/jetstream-announces-clearance-an-ai-zero-trust-reasoning-engine-1215883
