An AI-generated crypto transaction screenshot is not what it looks like: UK evidence rules after MoJ and Ayinde

Slug: ai-generated-crypto-transaction-screenshot-uk-evidence

Meta description: A screenshot is not proof of a crypto payment. See what UK evidence work after the MoJ review and Ayinde means for checking disputed records.

An AI-generated crypto transaction screenshot is not what it looks like: UK evidence rules after MoJ and Ayinde

An AI-generated or edited screenshot is an image of a claim, not proof that a crypto payment happened. In a dispute, the useful question is not simply whether the picture looks real. It is whether the underlying record can be checked, whether the person presenting it can explain its source, and whether the wider evidence supports the story.

That distinction matters in civil claims, criminal investigations, insolvency work and probate. A screenshot might be one item in an evidence bundle. It is not the blockchain record, a provider’s account record or proof that a named person controlled an address.

This is an educational guide, not legal advice. Court rules and professional guidance can change, and a live dispute needs advice from a suitably qualified professional.

What the MoJ review says about software-generated evidence

The Ministry of Justice’s 2025 Call for Evidence asked whether the long-standing approach to computer-generated evidence remains suitable for modern software. It points to the Post Office Horizon scandal as a warning against treating computer output as automatically reliable. The call covers software-generated evidence in criminal proceedings, including the way increasingly complex systems are used and checked.[2]

The call was a review, not a new rule saying that AI-generated material is automatically admissible or automatically rejected. It invited evidence about how the existing presumption should work in practice. Readers should therefore avoid treating the consultation as a ruling on any particular screenshot.[2]

The practical lesson is familiar: the person relying on a digital record needs to explain what it is, where it came from and why it should be trusted. A polished image does not answer those questions.

What Ayinde adds to the professional picture

In Ayinde v The London Borough of Haringey and Al-Haroun v Qatar National Bank, the Divisional Court dealt with serious problems involving the use of artificial intelligence in legal work. The judgment page records the neutral citation 2025 EWHC 1383 (Admin) and the associated case information.[3]

The Civil Justice Council’s 2026 interim report quotes the President of the King’s Bench Division in Ayinde: artificial intelligence is a tool with risks as well as opportunities, and its use needs appropriate oversight and compliance with professional and ethical standards.[1]

The same report proposes consultation on when legal representatives should declare AI use in preparing court documents. Its summary focuses on circumstances where AI has been used to generate evidence on which the court is being asked to rely. It distinguishes that from routine uses such as transcription, spelling checks and administrative assistance.[1]

This does not create a universal test for every screenshot. It does show the direction of professional concern: if software has helped create evidence, the people presenting it may need to understand and explain that use rather than treating the output as self-proving.

Why the crypto part is different

A crypto screenshot can display an address, amount, balance, timestamp, transaction hash or status label. Those details may help identify what should be checked. They remain details inside an image.

The underlying record may be held in more than one place:

  • a public blockchain record, if the transaction was broadcast and recorded there;
  • an exchange or payment-provider account record;
  • a wallet application or device record;
  • messages, invoices or other documents explaining the purpose of the transfer; and
  • records linking an address or account to a person or organisation.

These sources answer different questions. A blockchain may show that a transaction involving two addresses was recorded. It may not show who controlled either address, why the transaction was made, or whether the person in the screenshot is the account holder. An exchange record may identify a customer account, but it may also need to be checked for its creation, alteration and retention history.

The Civil Justice Council report discusses Crypto Open Patent Alliance v Craig Steven Wright in its treatment of AI-manipulated material, including concerns about hidden prompts, metadata and text inserted into documents. The point for a crypto reader is limited but important: even material that looks like a conventional document can require scrutiny of how it was produced and what was changed.[1][4]

A transaction hash is a starting point, not a verdict

If a screenshot includes a transaction hash, do not assume that the hash confirms the whole story. Check it on the correct network using an independent route. Confirm the asset, amount, sender and recipient fields, status, block or equivalent network position, and any contract or token details that affect the interpretation.

A matching hash can establish that a network record exists. It does not automatically establish:

  • who controlled the sending or receiving address;
  • that the person who supplied the hash initiated the transaction;
  • that the recipient is the company, exchange, charity or individual named in the message;
  • that the transfer was made for the stated invoice, loan, purchase or investment; or
  • that the funds are still available or can be recovered.

A genuine transaction can also be irrelevant to the claim. Someone could show a real transfer to an address they do not control, or a small payment while describing a larger payment elsewhere. Smart-contract activity may need technical interpretation rather than a simple reading of a wallet screen.

How to examine a suspicious screenshot

Preserve the original file and the message or email that accompanied it. Do not overwrite the original with annotations. Make a separate working copy if redaction is needed, and note when and where the material arrived.

Then work through the claim in this order:

  1. Identify the network. A Bitcoin transaction hash is not checked as an Ethereum transaction, and token transfers can depend on a particular network and contract.
  2. Capture the claim exactly. Record the stated amount, asset, address, hash, time, account name and purpose without silently correcting them.
  3. Check the underlying record independently. Type the identifier into a trusted route reached separately from the sender’s message. Do not rely only on a supplied link.
  4. Compare the complete record. Check the network, amount, asset or contract, sender, recipient, status and relevant time. A cropped or partial match is not enough.
  5. Check identity separately. A matching address does not prove that it belongs to the named person or organisation. Use a previously known contact route or separate provider record.
  6. Keep context with the technical record. An on-chain event does not explain the contract, invoice, loan or alleged investment on its own.
  7. Preserve sensitive information carefully. Do not publish recovery phrases, private keys, one-time codes, passwords, identity documents or unnecessary personal data.

If a request for money, a release fee or a recovery payment is attached to the screenshot, pause before sending anything. Do not connect a wallet or disclose a recovery phrase to prove a transaction.

What metadata and AI detectors can, and cannot, do

File metadata, edit history and provenance information can help explain how an image was created or handled. They are not conclusive on their own: metadata can be removed or changed, and a genuine screenshot can still show a misleading account, network or balance.

An AI detector has a similar limit. A result that says an image is probably synthetic is evidence about the image, not a finding that the crypto transaction did or did not happen. A result that says an image appears authentic does not prove the payment either.

The evidential question is therefore wider than detection. Compare the image with the underlying record and the surrounding documents. Record what each source can establish, and do not turn a confidence score into a conclusion about a person’s honesty or liability.

The legal-property frame is not the same as proof of a transaction

The Property (Digital Assets etc) Act 2025 says that a thing, including one digital or electronic in nature, is not prevented from being the object of personal property rights merely because it is neither a thing in possession nor a thing in action. The Act extends to England and Wales and Northern Ireland and came into force when it was passed.[5]

That statutory wording helps explain why cryptoassets can be treated as property in relevant legal work. It does not make a screenshot reliable, identify an address controller or settle a dispute about whether a payment occurred. Property status and evidential proof are separate questions.

What to keep for a professional review

For a significant dispute, keep the original screenshot, the original message, any supplied link, the full transaction identifier, the network name, account or wallet records, relevant correspondence and a simple chronology. Preserve rather than edit the original material. If personal or secret information is present, keep it secure and share only what an appropriate professional needs.

Do not publish a private person’s address or identity documents just because a screenshot is disputed. Do not accuse someone of fraud because an image looks unusual. The MoJ review and the court material discussed above point towards careful examination of digital evidence, not a shortcut based on appearance or an automated label.[1][2]

Related reading

Sources

[1] https://judiciary.uk/wp-content/uploads/2026/02/Interim-Report-and-Consultation-Use-of-AI-for-Preparing-Court-Documents-2.pdf — Civil Justice Council, Use of AI for Preparing Court Documents: Interim Report and Consultation
[2] https://www.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 — Ministry of Justice, Use of evidence generated by software in criminal proceedings: Call for Evidence
[3] https://www.judiciary.uk/judgments/ayinde-v-london-borough-of-haringey — Ayinde v The London Borough of Haringey and Al-Haroun v Qatar National Bank
[4] https://www.judiciary.uk/judgments/crypto-open-patent-alliance-v-craig-steven-wright — Crypto Open Patent Alliance v Craig Steven Wright and others
[5] https://www.legislation.gov.uk/ukpga/2025/29/enacted — Property (Digital Assets etc) Act 2025

Internal-link suggestions

  • Link the existing screenshot explainer to this article using anchor text such as “what a disputed screenshot can establish in evidence”.
  • Link the hash glossary from the verification checklist and the sentence explaining transaction identifiers.
  • Link the finality explainer where the draft distinguishes a recorded transaction from later settlement or reversal rules.
  • Link the existing privacy article from the warning that a public address can expose transaction history when linked to a person.

Leave a comment