Browser and operating system are one attack surface, not two. Salience Cyber instruments both and judges them in a single decision loop — at the point of interaction, before execution. A neuromorphic engine makes that decision in-process, so there is no kernel driver, no language model in the path, and nothing metered per token.
AIDR — AI Detection and Response. As the market practices it, the response is a ticket: the interaction is detected, recorded, and handed to someone. We do the other half. Inspection reads what an interaction is doing while it happens; enforcement acts on it before it executes. Most of the category stops at the first.
Enforcement runs on the endpoint and in the browser. AI traffic, agent actions and tool calls are inspected and acted on as they happen, with the System Sensor’s network, process and filesystem engines correlating them under one user, process and session identity. Discovery and the audit record come with it, because you cannot enforce on what you cannot see.
It runs alongside an existing EDR, XDR or endpoint protection platform — or several at once. Nothing to integrate, no console to replace, and no second kernel driver for change control to approve. The endpoint platforms sell AI capability as an add-on to their own estate, which means it arrives with their agent and their licence. This does not.
The sensors always run on the endpoint and always decide in-process. What you choose is where the Enterprise Console lives — on-premise, in your own private cloud, or in public cloud. For an organisation with a residency requirement, a classification boundary, or a regulator who asks where AI security data is held, that choice is the difference between a control you can adopt and one you cannot.
Thirty minutes with an engineer: your AI tools, your endpoints, and the decision loop running against them. No commitment, and nothing to uninstall afterwards.