The UK does not have a single public deepfake detector that can tell you whether a video, voice note or screenshot is genuine. The current public programmes do different jobs. The Information Commissioner’s Office (ICO) is setting expectations for responsible AI and auditing some uses of biometrics. The Financial Conduct Authority (FCA) is giving financial-services firms places, data and technical support to test AI. The Home Office is funding a wider counter-fraud system that includes law enforcement, data sharing and an Online Crime Centre.
That distinction matters if a crypto message arrives with a convincing voice, a familiar face or a screen recording of an alleged payment. These programmes may improve prevention, investigation and oversight. They do not make every piece of digital media provably authentic, and they do not guarantee that a victim will recover money.
What the ICO’s AI plan funds
The ICO’s “Preventing harm, promoting trust” plan of action is primarily a regulatory and assurance programme, not a consumer deepfake-detection service.[1]
Its 2025/26 work includes developing a statutory code of practice on AI and automated decision-making, sharing lessons from central-government use of automated decisions, auditing police forces that use facial-recognition technology, scrutinising foundation-model developers and examining accountability and redress issues around agentic AI.[1]
The March 2026 update says the ICO is preparing draft automated-decision-making guidance and work on an AI and ADM code of practice. It also says the regulator is engaging with 11 major AI foundation-model developers about data-protection compliance and measures to reduce harm.[2]
Those activities can affect how organisations collect data, explain automated decisions and protect people. They are not a promise that the ICO will authenticate a clip for an individual. Nor do they amount to a list of approved deepfake detectors.
The ICO’s technology research is useful for understanding the limits. Its synthetic-media report describes content generated partly or entirely with AI or machine learning, including images, video and audio. It discusses provenance and machine-readable marking, but also the privacy implications of automated moderation and the need to consider rights when decisions are made automatically.[3]
A provenance label can help a recipient understand where a file came from. It is not the same as proving that every claim in the file is true. An unlabelled file is not automatically fake either.
What the FCA AI Lab actually provides
The FCA’s AI Lab is aimed at UK financial markets. It is a route for the regulator, firms and other stakeholders to discuss AI use cases, risks and practical regulatory questions.[4]
The Lab includes the Supercharged Sandbox, AI Live Testing, AI Spotlight, the AI Sprint and the AI Input Zone.[4] The AI Sprint was a two-day event held in January 2025, with 115 participants from industry, academia, regulators, technology providers and consumer groups. Its published themes included regulatory clarity, trust and risk awareness, collaboration and safe innovation through sandboxing.[5]
That is research, engagement and supervision. It is not an enforcement unit and it is not a public hotline for checking whether a crypto video is real.
The FCA’s Supercharged Sandbox is closer to technical experimentation. It gives participating firms access to GPU-enabled computing, enterprise AI tools, synthetic or curated datasets, expert support and a controlled environment for testing.[6] The second cohort, which launched in July 2026, includes use cases intended to detect financial fraud and economic crime more effectively.[6]
The important word is “test”. A sandbox can help a firm evaluate a model, expose weaknesses and develop governance. It does not turn a prototype into a generally reliable deepfake detector. The FCA’s AI Lab pages do not say that the programme certifies a model, guarantees accuracy or approves a vendor for consumers.
The FCA’s work also sits inside financial regulation. It is concerned with risks to consumers and markets, including how firms use AI in areas such as compliance, fraud monitoring and customer support. That may matter when a cryptoasset firm uses automated systems, but it does not create a universal authenticity check for social-media content or private messages.
What the Home Office Fraud Strategy pays for
The Home Office Fraud Strategy 2026–2029 is the broadest of the three programmes. It says the Government will invest more than £250 million between 2026 and 2029 to deliver the strategy.[7]
The strategy describes deepfakes, large language models and voice cloning as technologies criminals are using to make attacks more convincing and harder to detect.[7] It does not say that the Government has solved that detection problem. In fact, it describes fraud controls as a continuing contest in which criminals adapt to new measures.[7]
The spending is spread across a counter-fraud system. One commitment is £31 million for an Online Crime Centre led by the Home Office and the National Crime Agency, working with policing, the intelligence community and private-sector partners. The strategy says the centre is intended to share data, analyse trends and coordinate interventions against online crime, initially focusing on fraud and high-volume cyber crime.[7]
Other measures include the expanded Stop! Think Fraud campaign, a stronger law-enforcement response, support for cyber-resilience centres, the Report Fraud service and planned work on victim support and justice.[7]
In practical terms, that is funding for capability: intelligence, coordination, prevention, reporting and enforcement. It is not a public grant scheme for buying a phone app, and the strategy does not name a commercial deepfake detector that readers can rely on.
Policy, enforcement and research are different things
It is tempting to describe all three programmes as “deepfake detection”. That loses the useful detail.
| Area | What the public material supports | What it does not establish |
|---|---|---|
| ICO | Guidance, regulatory codes, biometrics audits, foundation-model scrutiny and research on synthetic media | That the ICO authenticates individual videos or endorses a detection product |
| FCA | AI engagement, sandbox testing, live testing and financial-crime use cases in regulated financial services | That a sandbox model is certified, accurate in every setting or available to the public |
| Home Office and NCA | Counter-fraud investment, data sharing, online-crime coordination, reporting and enforcement | That every deepfake can be identified before a victim acts |
The Ministry of Justice’s 2025 call for evidence shows why the wider evidence problem remains open. It asked whether the common-law presumption that a computer was operating correctly remains fit for modern criminal proceedings. Its proposed scope included evidence generated by artificial intelligence and algorithms, as well as digital photographs, video footage, messages and extraction reports.[8]
That is a question about evidence and procedure, not a detection product. A disputed image may need its source, chain of custody, metadata, surrounding communications and other records examined together. A detector’s output, even when useful, is only one piece of that assessment.
What this means for a crypto scam or disputed transaction
If someone sends a deepfake video, voice note or screenshot connected with a crypto payment, treat the file as a lead rather than proof.
- Do not send crypto, recovery phrases, passwords or one-time codes because a familiar person appears on a call.
- Check the request through a separate channel that you already trust. Do not use the number, link or account supplied in the suspicious message.
- For a claimed on-chain payment, ask for the transaction ID and independently check the relevant public network record. A screenshot is not a transaction record.
- Keep the original message, URL, account name, timestamps and surrounding conversation. Avoid editing or re-saving files if they may later be needed as evidence.
- Contact your bank or payment provider promptly if money has moved, and report suspected fraud through the current UK reporting route. Recovery is not guaranteed, and speed does not make a report successful.
A detector may flag signs of manipulation. It may also miss a new technique or produce a result that needs human review. The safer habit is to verify the underlying event through an independent record and a second communication channel, rather than asking one tool to settle the whole question.
The short answer
The ICO’s programme funds regulatory guidance, assurance and research.[1]
The FCA AI Lab funds supervised experimentation and learning for financial-services AI, including financial-crime use cases.[4][6]
The Home Office strategy funds a wider counter-fraud response, including the Online Crime Centre, policing coordination, prevention and victim support.[7]
None of those sources promises a universal UK deepfake detector. They describe institutions building better rules, testing systems and coordinating responses to a changing fraud threat. Readers should expect more scrutiny and investigative capacity, not a guaranteed authenticity verdict for every digital file.
This article is general information, not legal, financial or fraud-recovery advice. If a case may become a legal dispute, speak to a suitably qualified professional.
Suggested internal links
- Crypto explained: start here — link from the first explanation of cryptoasset records and wallets.
- Existing article on AI-generated crypto transaction evidence — add once its final live URL is confirmed; link from the section on screenshots, transaction IDs and disputed evidence. Do not link to an unpublished draft.
- Existing blockchain transaction or hash explainer — add only after the exact live URL is checked; link from the on-chain verification example.
Sources
[1] ICO, “Our plan of action” — https://ico.org.uk/about-the-ico/our-information/our-strategies-and-plans/artificial-intelligence-and-biometrics-strategy/our-plan-of-action
[2] ICO, “AI and biometrics strategy update – March 2026” — https://ico.org.uk/about-the-ico/our-information/our-strategies-and-plans/artificial-intelligence-and-biometrics-strategy/ai-and-biometrics-strategy-update-march-2026
[3] ICO, “Synthetic media and its identification and detection” — https://ico.org.uk/about-the-ico/research-reports-impact-and-evaluation/research-and-reports/technology-and-innovation/tech-horizons-and-ico-tech-futures/tech-horizons-report-2025/synthetic-media-and-its-identification-and-detection
[4] FCA, “AI Lab” — https://www.fca.org.uk/firms/innovation/ai-lab
[5] FCA, “AI Sprint summary” — https://www.fca.org.uk/publications/techsprints/ai-sprint-summary
[6] FCA, “Supercharged Sandbox” — https://www.fca.org.uk/firms/innovation/supercharged-sandbox
[7] Home Office, “Fraud Strategy 2026-2029: Disrupting crime, supporting economic resilience and delivering justice” — https://assets.publishing.service.gov.uk/media/69ae77ddc78869bf8eb8a509/fraud-strategy-web.pdf
[8] Ministry of Justice, “Use of evidence generated by software in criminal proceedings: Call for Evidence” — https://gov.uk/government/calls-for-evidence/use-of-evidence-generated-by-software-in-criminal-proceedings/use-of-evidence-generated-by-software-in-criminal-proceedings-call-for-evidence