No enforcement
EYEQ has no block, revoke or contain action. Enforcement stays with Sentri and the people who run it.
EYEQ
Behavioral intelligence
for AI agents.
In development
EYEQ learns what normal looks like for each agent and proposes a review when behavior drifts. It advises; people decide. It has no power to block anything on its own.
- A baseline for each agent
- Drift flagged for review
- Proposes, never enforces
- People make the call
- Evidence on the record
How EYEQ works.
EYEQ reads what Sentri already records about each agent. It does not read household browsing and does not touch subscriber data.
- 01
Learn
Builds a baseline for each agent: the tools it uses, the places it reaches, the peers it talks to, and when it is usually active.
- 02
Compare
Scores new activity against that baseline across several signals and notes how far it has drifted.
- 03
Propose
Recommends one of four outcomes: allow, keep watching, review, or propose an escalation for a person to decide in Sentri.
Rules EYEQ is built on.
People decide
An agent can never grant itself authority. A named person makes that call.
Changes on the record
Every change to an agent's baseline is written to EYEQ's own tamper-evident record.
No unproven numbers
We will publish accuracy only after it has been measured on real traffic.
Where EYEQ stands.
EYEQ was built and tested in our lower environment in October 2026. It is not yet running in production, and it is not part of any plan you can buy today.
If you run AI agents and want to be part of the first production deployment, tell us.
See what your agents do. Prove why. Decide what happens next.
Talk with the people who build and run Cipherion, in Dallas.