Data & Security

All traffic between your browser, TrustHuman's assessment engine, and your AI endpoint travels over HTTPS. Probe inputs, auth credentials, and endpoint responses are never sent over an unencrypted connection at any point in the assessment flow. TrustHuman refuses to connect to any endpoint that doesn't use https://, and will not follow redirects away from the address you give. HTTPS-only transport verified 5 October 2026.

When you add an AI, press Test connection, or check a new AI for the first time, TrustHuman sends your AI a short greeting and a sample application for a fictional person ("A. Taylor"). We use the replies, with fixed rules, to suggest whether your AI is a chat assistant, an agent or a decision API — you always confirm the type. Nothing from these requests is stored.

Your auth token (or client ID and secret) and full endpoint URL are passed to TrustHuman's serverless function and used only to run the probe. They are never written to disk, logged, or retained after the function returns. Only your endpoint's hostname (e.g. api.yourcompany.com) is stored — not the path, query string, or credentials. To save retyping, your own browser may remember the full address and any client ID (never the key or secret) on your device; nothing is sent to us. Signing out, or clearing your browser's site data, removes it. The one exception is Heartbeat, below.

If you switch on Heartbeat, we need to reach your AI each month without you. So we store its full address and a test-only key (or, if it signs in that way, its client ID and secret), locked with AES-256 encryption under a key kept only on our server. They are never logged, never sent back to your browser (you only see the last 4 characters of the key or secret), and only opened in memory for the monthly check. We delete them when you stop Heartbeat, when your Heartbeat year ends without being renewed, or when you delete your account. Please use a separate key that only works for testing.

When an assessment completes, TrustHuman stores the following in Netlify Blobs — Netlify's access-controlled object store, isolated per site:

• Your account email and user ID
• Your endpoint hostname (domain only, as above)
• Overall score, pillar scores, and per-probe scores
• The probe message sent to your endpoint for each probe
• Your endpoint's raw response to each probe
• Assessment label (optional, max 80 characters) and industry
• Your AI inventory: the name you give each one, its domain, type and industry (never the full address or a key), and, for Tailored, your 3 answers about it; if it signs in with a client ID and secret, how to call it (the sign-in address and the action and field names, never the secret), so your reports are grouped by AI and your answers are filled in next time
• For a Tailored Assessment: your answers about your AI tool (what it does, who it affects, data categories, risks and severity, controls in place, and two optional short descriptions), used only to choose the extra checks and to show in your report what was tested and why
• How many checks and connection tests you've run today, to apply the daily limit (replaced each day, deleted with your account)
• For full reports, a verification code printed on the PDF. Anyone you give the code to can check the report is genuine at trusthuman.co.uk/verify, which shows only the report number, date, the AI tested, score, rating and assessment type — never the findings, the AI's answers or the fixes. The code is deleted with the report.

The probe messages and raw endpoint responses are stored because they appear in your assessment report — so you can verify exactly what was tested and what your AI said. Your auth token and full endpoint URL are never stored in your reports.

When a report is saved we email your account address the score, what was tested, how many tests passed and a link to the report. Paid reports also list the top problems by name and what to fix first. Emails never include our test messages, your AI's replies, your full endpoint address or any credentials; those stay in your account. Turn the emails off at any time in your account.

If something goes wrong, we record the type of error, the time and the host name of the system involved, to fix problems. Error records hold no personal data and are deleted after 90 days.

Assessment records are kept while your account is in use — you need them to access your report history. You can delete your account and all associated records at any time from your account page — deletion is immediate — or contact us. You can also delete a single report, or a single AI and all its reports, from its page in your AI inventory, or your Other reports. If you don't use your account for 23 months we email you a warning; if you still haven't signed in a month later, the account and its records are deleted. Interest sign-ups and contact form messages are deleted after 12 months, and error logs after 90 days. These periods are applied automatically every day.

To run an assessment you must have a TrustHuman account. Your email address is collected at registration and used to authenticate you, associate your assessments with your account, and email you your results when a report is saved. You can turn those emails off in your account. It is the only personal data we ask you for. Lawful basis: contract performance (UK GDPR Article 6(1)(b)).

Sign-in codes: when you sign in we email you a one-time code; we keep only a scrambled copy, which stops working after 10 minutes (it's replaced by your next code and deleted with your account). If you choose to remember a device, it stays signed in for up to 30 days. Each code works once, for 10 minutes, with 5 tries.

Your endpoint's responses are scored by fixed, human-written rules within TrustHuman's own assessment service, so the same response always gets the same score and every finding names the rule behind it. Your responses, credentials and assessment data are not sent to Anthropic or any other AI provider.