The trust stack for governed AI

Safe identity. Reliable AI.
Accountable governance.

We engineer the architecture that makes AI safe to use, reliable in operation, and accountable for what it does, cryptographic, per-inference, and architecturally unforgeable.

Safe · Reliable · Accountable / 10 UKIPO patent applications / IBM reference build partner
Our working paper · Published 12 May 2026 · CC BY 4.0

Read our published paper: Closing the EU AI Act Article 12 logging gap

A technical specification for how aiGEC™ closes the per-inference governance evidence gap that high-risk AI systems must satisfy under the EU AI Act, with application timelines subject to the final adopted AI Omnibus amendments. Prepared for engagement with BSI ART/1 and the appropriate ISO/IEC JTC 1/SC 42 channels.

Read paper on Zenodo →

Existing systems are strong at recording what AI generated. They generally do not record whether it was governed.

Content provenance standards (C2PA) record the origin of AI-generated content. Trusted execution environments certify the hardware. Governance platforms log aggregate activity. None of them, individually or in combination, issues a per-inference, externally-verifiable certificate proving that a specific AI output was subject to a non-bypassable governance evaluation before delivery.

Regulators, banks, and healthcare bodies are starting to ask the question existing systems can't answer. We've engineered the architecture and the certificate that does, and we hold ten UK patent applications directed to it.

One request. Three evaluations. One verifiable outcome.

Every governed AI interaction passes through three architectural checkpoints. The identity layer reduces footprint before the model sees the request. The governance layer evaluates the candidate output against runtime parameters before delivery. The evidence layer issues a cryptographically signed certificate as a structural output of that evaluation. None of the three can be bypassed.

Input
User request
Unverified
Layer 01 · Identity
synapse-ID
Footprint reduced
Safe
Layer 02 · Governance
aiGUARD
Decision evaluated
Reliable
CONFIDENCE CONSEQUENCE USER STATE
Layer 03 · Evidence
aiGEC
Certificate signed
Accountable
Output
Certified output
Verifiable
01 Safe Identity footprint reduced before the model sees the request
02 Reliable Confidence, consequence, and user state evaluated before delivery
03 Accountable Cryptographic certificate issued, externally verifiable, architecturally unforgeable

A patent-application portfolio.

10
UKIPO patent applications
3
Patent-pending products
4
Regulated verticals
1
Reference build partner

Maps to the frameworks that matter.

The trust stack maps directly onto the live regulatory frameworks for high-risk AI, from EU AI Act high-risk obligations, with application timelines subject to the final adopted AI Omnibus amendments, through FCA Consumer Duty, to public-sector accessibility requirements.

EU AI Act Art. 9 Risk Management EU AI Act Art. 12 Logging EU AI Act Art. 13 Transparency FCA Consumer Duty NHS Patient Safety MHRA · CQC WCAG 2.2 Accessibility Public Sector Accessibility Regs

Building with global infrastructure.

Reference build partner

IBM

IBM is our reference build partner for the aiGUARD™ architecture. SoW1 signed February 2026; SoW2 production build in scoping. The reference build establishes the integration pattern for enterprise agentic AI deployments across regulated industries.

Standards engagement

Mastercard / Google · Verifiable Intent

aiGUARD Systems monitors the Verifiable Intent agent_attestation specification, a published draft open specification for agentic commerce trust. aiGEC™ is positioned to populate the agent_attestation field with cryptographically verifiable governance proof.

Four verticals. One commitment: safe, reliable, accountable AI.

Regulated financial services

Agentic AI advice with verifiable governance, mapped to FCA Consumer Duty obligations on vulnerable customer protection.

NHS and healthcare AI

Per-output certification for patient-facing systems, retrievable by output ID and verifiable by CQC, MHRA, or NHS England without internal system access.

Enterprise agentic AI

Provable governance across multi-system pipelines, with cryptographic chain of custody from agent to output.

Public sector accessibility

A proposed per-inference certified accessibility compliance field for standards engagement, supporting Public Sector Bodies Accessibility Regulations.

Verify, don't assume.

If you regulate, build, or buy AI systems where someone needs to verify, not assume, that an output was governed and a person's footprint wasn't quietly expanded in the process, please get in touch.

Contact us