Certification Standards

Trust Tier Requirements

Every GAMP tier has specific, transparent requirements. No ambiguity. No hidden criteria. Here is exactly what you need to achieve each tier and what it means to the people who verify your model.

● Declared
Tier 1 — Declared
The entry point. You have declared what your model is, who built it, and what it is used for. GAMP issues a cryptographic passport binding your identity to your declaration. No independent verification yet — but the declaration is permanent and signed.
⏱ Registration takes 10 minutes · Free during launch
Base score: 10
Required to register
Model name and version number
Model type (LLM, classifier, multimodal, etc.)
Intended use — what problem does it solve?
Known limitations — what should it not be used for?
Training data description (can be high-level)
License type
What you receive
Ed25519 cryptographic passport — permanently signed
Unique model ID and passport hash
Live trust score (starts at 10)
Embeddable trust badge for your website
Public trust certificate URL — shareable with anyone
Public transparency log entry
● Verified
Tier 2 — Verified
Your declaration has been independently reviewed and cryptographically signed evidence has been submitted. GAMP has confirmed that what you declared matches what you provided. Your badge turns green and your trust score jumps to a minimum of 40.
⏱ Review typically 2–5 business days after document submission
Base score: 40
Required for verification
All Declared tier requirements
Architecture documentation — how the model is built
Training data evidence — source, scope, and licensing
At least one Verifiable Evidence Packet (VEP) submitted and verified
Intended use verified against architecture and training data
Additional benefits
Green verified badge — visible trust signal on any website
Trust score minimum 40 — strong signal for procurement
VEP evidence chain — cryptographic proof on record
Eligible for cosignatures from GAMP certifiers
Downloadable compliance certificate PDF
● Audited
Tier 3 — Audited
An independent GAMP-accredited certifier has co-signed your passport. This is independent third-party validation — not just GAMP's review. Required for high-risk AI deployments under the EU AI Act and equivalent frameworks. Trust score minimum 70.
⏱ Certifier engagement — timeline varies by certifier
Base score: 70
Required for audit
All Verified tier requirements
Safety evaluation documentation
Red-team or adversarial testing report
Bias and fairness assessment
Co-signature from at least one GAMP-accredited certifier
Regulatory alignment mapping (EU AI Act, NIST, ISO 42001)
What Audited unlocks
Blue audited badge — enterprise-grade trust signal
Certifier co-signature on public passport
Regulatory evidence package for EU AI Act conformity
Suitable for procurement requirements at financial institutions and healthcare
ISO/IEC 42001 alignment documentation
● Accredited
Tier 4 — Accredited
The highest GAMP certification tier. Multiple independent certifiers have co-signed your passport, a full conformity assessment has been completed, and ongoing monitoring is in place. This is the standard for mission-critical AI in regulated industries.
⏱ Full accreditation process — 4–12 weeks depending on model complexity
Base score: 90
Required for accreditation
All Audited tier requirements
Co-signatures from minimum two independent GAMP certifiers
Full conformity assessment against applicable regulatory framework
Continuous drift monitoring actively configured
Incident response plan documented
Annual re-certification commitment
What Accredited means
Gold accredited badge — highest visible trust signal
Multiple certifier co-signatures — chain of independent trust
Suitable for government, defence, and critical infrastructure deployments
Full EU AI Act high-risk system conformity evidence
Priority support from GAMP compliance team

How the trust score is calculated

Every score component is transparent and publicly visible on each model's trust certificate.

Component Declared Verified Audited Accredited
Base tier score+10+40+70+90
Each certifier co-signature+3+3+3
Co-signature bonus cap+10 max+10 max+10 max
Verified within 30 days+5+5+5
Verified VEP penalty claim−12−12−12−12
Unverified drift claim−4−4−4−4
Critical unresolved drift−7.5−7.5−7.5−7.5
Flagged for review−20−20−20−20
Suspended0 (floor)0 (floor)0 (floor)0 (floor)
Score range0–1000–1000–1000–100
Frequently asked questions
Can I register for free? +
Yes. Registration is free during the GAMP launch period. All four tiers are available at no cost until 20 June 2026. After that, Verified, Audited, and Accredited tiers will carry an annual fee. Declared tier remains free permanently.
How long does verification take? +
Declared tier is instant — your passport is issued the moment you register. Verified tier review typically takes 2–5 business days after document submission. Audited and Accredited tiers depend on certifier availability and model complexity.
Do I need to share proprietary model weights? +
No. GAMP does not require access to your model weights, training data, or proprietary code. You declare what your model is and submit documentation describing it. The cryptographic passport binds your identity to that declaration — not to the underlying implementation.
What counts as a Verifiable Evidence Packet? +
A VEP is a cryptographically signed document attached to a specific claim about your model. It can be a safety evaluation, training data audit, architecture specification, compliance assessment, or drift detection report. Any document that independently supports a claim about your model's behaviour or composition qualifies. The VEP is signed with your private key and cannot be altered after submission.
How does GAMP map to EU AI Act requirements? +
GAMP Audited tier maps to the documentation and conformity assessment requirements for high-risk AI systems under the EU AI Act. The VEP evidence chain provides the technical documentation layer. The transparency log satisfies ongoing monitoring requirements. GAMP Accredited tier is designed to satisfy the full conformity assessment requirements with multiple independent certifier co-signatures.
What happens if my model's behaviour changes? +
GAMP's semantic drift detection continuously monitors deployed model outputs against the registered baseline. If significant drift is detected — meaning the model's behaviour no longer matches what was declared — a drift report is automatically raised and the trust score updates immediately. Your badge reflects the current state in real time. You are notified via webhook and email when drift is detected.

Start with Declared tier — free, 10 minutes

Register your model today. Upgrade to higher tiers as your compliance requirements evolve.

Register Your AI Model →