01
Verify before consequence
Trust checks belong before a sensitive action, not only in an audit after the damage is done.
About Agent Trust Gate
Building trust at the point AI becomes action.
Agent Trust Gate™ is an independent UK technology project focused on a narrow but increasingly important question: how can an organisation verify authority for one consequential AI-agent action before it happens?
Our focus
Close to the action. Close to the evidence.
ATG does not try to become an entire AI governance platform. It concentrates on the pre-action boundary where a system must decide whether this requester, under this authority, may perform this exact consequential action now.
That focus has shaped the public demonstrator around Agent Standing, Verified Human Authority, exact-action GatePass, refusal behaviour and separate decision/execution evidence.
Mission
Make digital trust verifiable before action, observable during execution and accountable afterwards.
Control before action. Evidence during execution. Accountability after the event.
Operating principles
The way we build matters.
The public project deliberately favours clear boundaries and inspectable evidence over inflated capability claims.
02
Fail closed when proof is unresolved
Unverifiable standing, stale proof, changed action or missing authority should not silently degrade into permission.
03
Keep humans accountable
Where policy requires human authority, the human must be identifiable, current and authorised for the exact action.
04
Make evidence inspectable
Technical reviewers should be able to reproduce core demo behaviour and see refusal paths in public.
05
Separate decision from execution
An allowed action is not automatically an executed action. Different receipts should record different facts.
06
Stay honest about boundaries
Local demonstrator, production system, certification and commercial deployment are different stages and must not be conflated.
Founder & project lead
Gareth Price
Agent Trust Gate is led from the United Kingdom by Gareth Price as Founder and Project Lead.
The project is intentionally public at the demonstrator level so serious reviewers can inspect the architecture, reproduce technical paths and challenge the claims. Commercial production integration, if pursued, would be separately scoped and secured.
Current status
Public local demonstrator
- Working deterministic technical demonstrations
- Public reviewer and evidence paths
- Corporate technical-enquiry route
- Bounded paid evaluation pilot available by written scope
- No production deployment or certification claimed
Future-ready without future theatre
Designed not to be trapped by today’s assumptions.
ATG’s roadmap considers stronger agent identity, machine-to-machine trust, crypto agility, post-quantum readiness and low-latency risk-tiered gating — while keeping today’s public implementation claims firmly tied to what reviewers can inspect now.
Human-readable trust proof
Technical receipts should ultimately be understandable to customers, operators and accountable decision-makers.
Crypto agility
Proof formats should be able to evolve as algorithms, credentials and evidence standards change.
Fast reaction trust
Low-risk actions may eventually need safe fast paths, while higher-risk actions remain gated or refused.
Serious review welcome
Contact Agent Trust Gate →Challenge the architecture. Bring a bounded scenario.
We welcome technical criticism, reviewer feedback and design-partner conversations that respect the public project’s safety boundaries.