UK GDPR
Lawful basis for training and inference data, DPIA triggers for AI, the Information Commission (it replaced the ICO on 30 September 2026), and what a controller or processor must hold.
Are you a provider or a deployer? Which risk tier? Which dates apply to you, as the European Commission states them? And the ten questions enterprise buyers now ask AI vendors. Written for UK founders and CTOs with 5 to 50 people.
Free download
Lawful basis for training and inference data, DPIA triggers for AI, the Information Commission (it replaced the ICO on 30 September 2026), and what a controller or processor must hold.
Provider or deployer. Four risk levels. It applies to UK firms placing systems on the EU market or whose output is used in the EU. The dates, as the Commission states them.
The AI management system standard: 38 Annex A controls, how it sits on ISO 27001 or SOC 2, and when a buyer asks for it.
Which AI systems run in your product and what data do they touch?
Keep an AI system register: every model, API and tool, with data touched and owner.
Are you a provider or a deployer under the EU AI Act, and for which systems?
Provider: you build or brand the system; deployer: you use someone else's system in your product or operations, and you can be both.
Which risk tier, and on what basis?
The Act has four risk levels: unacceptable (banned), high (Annex I and III), limited (transparency duties) and minimal.
Do you keep an AI system register and an AI risk register?
The register comes first, then an AI policy, risk register and named owner.
Who is accountable for AI governance, by name?
Buyers expect a named owner for the AI policy and risk register.
Do you use customer data to train or fine tune models?
Under UK GDPR you need a lawful basis per purpose for training and inference data.
How are model outputs monitored, and who reviews them?
Buyers ask this inside enterprise security reviews of AI startups, so name the reviewer and the method in writing.
Which third-party models and APIs sit underneath, and under what terms?
Hold processor terms with every model and hosting supplier, and have your sub-processor list ready.
How do users know they are interacting with AI?
Limited-risk systems carry transparency duties, and general application, including transparency rules, starts on 2 Aug 2026.
Are you working towards ISO 42001, and what evidence can you show today?
ISO/IEC 42001 is the certifiable AI management system standard, and certification is issued by an accredited certification body, never by your consultant.
Questions 1, 5 and 7 follow the pattern safet.com reports in enterprise security reviews of AI startups; the rest follow the Act's own provider and deployer duties.
Regulation (EU) 2024/1689 entered into force on 1 August 2024. The Annex III and Annex I dates follow the Digital Omnibus on AI, in force 27 July 2026. Source: digital-strategy.ec.europa.eu, read 31 October 2026.
A free 45 minute diagnostic with Alfred Obeng, CISSP, ISO 27001 and ISO 42001 Lead Auditor. You leave with a one page plan the same day.
Book the free diagnosticWe use cookies to measure the site and improve it.
We use cookies and similar technologies to measure how this site is used and to measure our advertising. Until you choose, Google tags run in a consent-limited mode that sets no cookies. If you accept, we load Google Analytics 4 and Google Ads with cookies, PostHog, Plausible and Microsoft Clarity, which records session replays. Read our Cookies policy.