Can digital forensics show an AI-generated image was made on a particular device?
On 26 September 2026 Hong Kong’s Secretary for Justice said the Law Reform Commission should consider whether making non-consensual AI-generated sexual images should itself be an offence, even if they are never shared. If the law moves that way, cases will turn on a forensic question: can you show that a particular image was made on a particular device, by a particular person?
Sometimes, to a degree. Examining a phone or computer can show AI apps and accounts, prompts and generation history, source photographs, the files produced and their timestamps, and whether anything was shared or deleted. Cloud accounts may hold more. What forensics usually cannot do alone is prove who was holding the device; that comes from the wider evidence. The image file on its own, especially after being shared, often says little about where it was made.
What was reported
RTHK reported that Secretary for Justice Paul Lam said whether making non-consensual AI sexual images should be an offence, even without distribution, “must be one of the research topics”, and that no conclusion had been reached. He described several levels to the problem: production, distribution, and use for further offences such as extortion. The AI element has been separated from the wider review of computer crime so that it can move more quickly to public consultation.
What a device can show
- Apps and accounts. Which image-generation apps or services were installed or used, and under which accounts.
- Inputs. Source photographs of the person depicted, and in some cases the prompts or settings used.
- Outputs. Generated files, their creation times and any embedded metadata or provenance marks.
- Movement. Whether images were sent, uploaded or shared, and to whom.
- Deletion. Traces of files, apps or histories that were deleted, which may be relevant in themselves.
The limits
Many services generate images in the cloud, so the device may hold only the result and an account login; the service provider may hold the rest, and obtaining it may require legal process. Shared images are often stripped of metadata by messaging apps. And a device shows what happened on it, not who was holding it, so attribution depends on the wider evidence. A careful expert report says which of these points the evidence supports and which it does not.
If you are a victim
Do not download, save, screenshot or forward the images. Write down where they appeared, with dates, links and account names, taking a screenshot only where no image appears on the screen, and report it to the police, who will collect the material. If the person shown may be under 18, report to the police straight away and leave the phone and any examination to them; do not have it copied privately. Tell someone you trust. Do not pay or reply to anyone demanding money. Once you have recorded the links and account names and reported it, ask platforms to remove the images. Where a civil claim or prosecution follows, the original messages and devices are the evidence, so do not delete them.
See mobile phone forensics and computer forensics.
Sources
General commentary on public reports, not legal advice and not a finding about any organisation or person named in those reports. Facts are as reported at the time of writing.
Frequently asked questions
Can digital forensics show an AI-generated image was made on a particular device?
Sometimes, to a degree: a device can show AI apps and accounts, source photos, generation history, output files and timestamps, and sharing or deletion. Where generation happened in the cloud, the provider may hold key records. Proving who was holding the device depends on the wider evidence.
Is making a deepfake sexual image a crime in Hong Kong?
Publishing or threatening to publish intimate images without consent, including altered images, can already be an offence under section 159AAE of the Crimes Ordinance. In September 2026 the Secretary for Justice said whether creating them should itself be an offence was a topic for law reform research, with no conclusion reached.
Does an AI image contain metadata showing who made it?
Sometimes it contains metadata or provenance marks from the tool used, but these are often stripped when images are shared through messaging apps or social media.
What should a victim of a deepfake image do?
Do not forward or download the images. Record where and when they appeared and who sent them, report to the police, who will collect the material, do not pay anyone demanding money, and ask platforms to remove the images. If the person shown may be under 18, do not save copies at all.
Can deleted AI images be recovered from a phone?
Sometimes, depending on the device, how long ago they were deleted and whether they were synced to cloud storage. Evidence of the deletion itself may also be relevant.
Facing something like this?
The first discussion is confidential and free, and usually changes what you decide to do next. Call +852 5808 1071 or use the case review form.