Your AI is live.
Is it fair, transparent and lawful?
Independent tests of your real AI against UK and EU law — scored by fixed, human-written rules, not another AI.
Free TrustScore · Free account required
AI regulation is reviewed monthly against official sources — including the ICO, gov.uk, and the EU Commission — and changes are applied only after human review. How we stay current →A 30-second animation of three short chats with a bank's AI assistant. First, asked "am I talking to a real person?", it never says it is an AI (EU AI Act Article 50). Next, Amara Eze, a first-time buyer, asks what mortgages are on offer and is told certain factors may affect eligibility, while Olivia Bennett asks the same question and is offered a mortgage adviser (Equality Act 2010). Last, asked for a copy of their data, it says up to 90 days, when the legal limit is one month (UK GDPR Article 12(3)). TrustHuman tests your live AI product and scores every reply with fixed, human-written rules, not another AI. The examples are illustrative.
Aligned to major regulatory frameworks
Three questions every
AI deployment needs to answer
AI tools now make or shape decisions about real people. Regulators expect you to show that yours treats people fairly, explains itself and protects their data.
-
01Is this AI biased?AI trained on historical data can repeat historical bias. We send matched pairs of requests that differ in one thing — such as age or a name — and check whether your AI treats them differently.Equality Act 2010 · EU AI Act Annex III
-
02Can it explain itself?People have a right to a meaningful explanation of automated decisions, and to know when they're dealing with an AI. We check whether your AI gives specific reasons, not vague policy references.UK GDPR Article 22C · EU AI Act Article 50 · ICO guidance
-
03Is data being protected?We check whether your AI can be tricked into revealing internal information, and whether it handles data rights requests correctly and on time.UK GDPR · ICO guidance on AI and data protection · EU AI Act Article 15
Real tests on your real AI.
Not questionnaires.
-
Step 1Connect your AIPaste your AI tool's web address. We check the connection and suggest what type of AI it is. No code changes on your side.
-
Step 2We test itWe send a fixed set of realistic test requests using made-up people — never your customers' data — and score every reply with fixed, published rules.
-
Step 3You get a reportYour TrustHuman Score, pillar scores and a Bronze, Silver or Gold rating, with each finding mapped to the law it relates to. We don't just flag issues. The full report tells you what to fix for each one.
Always assessed against current law
AI regulation is moving fast. Our scoring rules are kept in step with it, and every change is checked by a person before it counts.
Three assessments.
One trusted platform.
Try the free demo to see how it works, get your free TrustScore, then commission a full report when you're ready.
Built to be checked, not just trusted
Free TrustScore · Free account required
What people ask before they assess
What is TrustHuman?
TrustHuman is an independent AI ethics assessment platform. We connect to your AI tool's API endpoint, send carefully designed synthetic test inputs called probes, and score the responses against specific legal and ethical criteria. You receive a TrustHuman Score, a detailed findings report, and a Bronze, Silver, or Gold rating. We are not a consultancy and not a questionnaire — we probe your AI directly to see how it actually behaves.
What does TrustHuman test?
TrustHuman tests across three compliance pillars: Bias and Discrimination (40% weighting) — whether the AI treats people differently based on age, postcode, or socioeconomic background, mapped to Equality Act 2010 and EU AI Act Annex III; Transparency and Explainability (35% weighting) — whether the AI discloses it is an AI and can explain its decisions, mapped to EU AI Act Article 50 and UK GDPR Article 22C; Data Privacy and Security (25% weighting) — whether the AI is vulnerable to prompt injection and handles data subject rights correctly, mapped to UK GDPR and the ICO guidance on AI and data protection.
What does the TrustHuman Score mean?
The TrustHuman Score is a weighted average of your pillar scores, from 0 to 100. A higher score means better alignment with legal and ethical requirements. 90–100 is Gold (excellent alignment, continue monitoring); 75–89 is Silver (good alignment with minor gaps, address priority findings); 60–74 is Bronze (adequate, material gaps present, supervised deployment recommended); below 60 is Not Certified (significant compliance failures, immediate remediation required).
When should I use TrustHuman — before or after deploying my AI?
Both — and in between. You can run an assessment before you deploy a tool to check what you're about to put live, on your live system to see how the AI you're actually running behaves, and again after any change — a new data source, a model update, a new integration. Compliance is not a one-time event; the AI running today is not always the one you checked last month.
Is TrustHuman free to use?
Yes — TrustHuman offers a free TrustScore. Create a free account, connect your live AI tool's endpoint, run a full assessment, and receive your overall TrustHuman Score with a traffic light rating at no charge. The free TrustScore shows your overall score and rating, your pillar scores, and which checks passed or failed and how serious each failure is. The full Snapshot Assessment report adds the findings behind each result, the evidence (what we sent and how your AI replied), the exact legal provisions, and recommended actions.
How often should we reassess?
AI tools change — models are updated by vendors, prompts and configurations drift. A tool that passed an assessment six months ago may behave differently today. Our guidance is a full reassessment every six months, or whenever the underlying model, configuration, or deployment context changes significantly. The Heartbeat Assessment re-runs your Snapshot or Tailored Assessment every month and emails you each result, so you can see how results change over time.
What data does TrustHuman retain?
All probes use synthetic inputs only — no real customer data is used or transmitted at any point during an assessment. Your API credentials are used only to make the probe calls and are not stored by TrustHuman, unless you switch on Heartbeat: then a test-only key is stored locked (encrypted) for the monthly checks and deleted when Heartbeat stops. Your saved assessment reports, including the probe messages and your AI's responses, are kept in your account so you can review them. They're deleted on request (including when you delete a report, an AI or your Other reports in your AI inventory, or your account) or after 2 years of inactivity. When a report is saved, we email you your results: the score, what was tested and a link to the report. Emails never include the probe messages or your AI's responses, and you can turn them off in your account.
Can TrustHuman assess any AI tool?
TrustHuman can assess any AI tool that is accessible via a REST API endpoint, which covers the vast majority of deployed customer-facing AI tools. You need to provide an endpoint URL and an authentication method. For chat and agent tools, TrustHuman sends each test as JSON in a message field and reads the reply from a response, message, text, content or answer field. For decision APIs, it sends standard test applicants (name, age and postcode); mapping to your own field names isn't available yet. Tools that cannot currently be assessed include AI tools with no API access, tools that require browser-based interaction only, or tools behind enterprise firewalls without external access.
Which regulations does TrustHuman map to?
Every finding in a TrustHuman report maps to a specific legal obligation: Equality Act 2010 (bias and discrimination findings), UK GDPR Article 22C (safeguards for automated decisions: information, human review and the right to contest), EU AI Act Article 50 (AI disclosure obligations), EU AI Act Annex III (high-risk AI system requirements), ICO guidance on AI and data protection (data privacy and security findings), and UK GDPR (data subject rights handling).
Will a TrustHuman certification satisfy our regulator?
TrustHuman certifications are evidence of due diligence, not a substitute for regulatory compliance. A TrustHuman report provides documented, evidence-based proof that you assessed your AI tool against specific regulatory criteria, identified findings, and were in a position to act on them. Regulators look favourably on organisations that demonstrate active and structured governance of their AI systems. We are not lawyers and this is not legal advice.
Want to partner with TrustHuman?
TrustHuman partners with AI platforms, marketplaces, accountants, advisors, and business services to bring independent compliance assessment to their customers. Non-exclusive, transparent, and built around your model.