Provenance Gate
A security mechanism for AI agent memory writes that requires facts proposed by an agent to specify a source before they can be written to memory. It performs four categories of verification: junk and duplicate checks, contradiction checks, a confidence floor, and provenance checks.
Timeline (last 90 days)
The first version of the provenance gate was fully breached in 6 attacks, and after the fix all 30 attacks were blocked.
The provenance gate performs four types of checks: junk and duplicate checks, contradiction checks, a confidence floor, and provenance checks.
In a 30-attack test, the gate let through 9 of 10 legitimate customer confirmation updates, and the one it missed was a disclosed race condition.
All Facts (3)
The provenance gate performs four types of checks: junk and duplicate checks, contradiction checks, a confidence floor, and provenance checks.
50%UnverifiedThe first version of the provenance gate was fully breached in 6 attacks, and after the fix all 30 attacks were blocked.
50%UnverifiedIn a 30-attack test, the gate let through 9 of 10 legitimate customer confirmation updates, and the one it missed was a disclosed race condition.
50%