UODO AI Deepfake Case: Public Photo Does Not Mean Consent

Status as of: October 10, 2026

The Polish Personal Data Protection Office (UODO) detailed an official notice sent to the police commander in Kraków following media reports that a minor student’s photo was altered by another minor using AI. The announcement from June 24, 2026 concerns an intervention, not a judicial ruling. The crux of this case lies in the line between technical access to an image and the legal right to use it. The fact that an image is fabricated does not erase the question of liability for its creation and distribution.

Timeline

1. June 24, 2026 - Regulatory announcement regarding the intervention.
2. September 22, 2026 - Broader policy statement on deepfakes and illegal content, not a formal resolution of the Kraków case.

What the Actions Indicate

From Media Reports to Regulatory Response

The starting point was media reporting. UODO emphasized that the public availability of a photograph does not automatically establish a legal basis for processing it in this manner. UODO, June 24, 2026. This is a critical distinction: the ability to download a file describes technical access, not the legality of repurposing it. The question is not merely where an image came from, but what is done with it.

Fake Image, Real Responsibility

Authenticity and harm require separate assessments. Establishing that an image has been manipulated answers whether it reflects reality. It does not answer whether co-opting someone’s likeness violates their rights. An imposed context, loss of control over personal representation, and exposure to outside judgment remain key criteria even when viewers recognize the image as synthetic. An image does not need to be perceived as authentic for the violation to be real.

The argument based on public availability shifts accountability to the wrong party. It implies that anyone posting an image must foresee every malicious manipulation, while the person using the generative tool faces no scrutiny. A sounder principle focuses on the act itself: the purpose, scope, and distribution channel. Caution exercised by the person depicted does not substitute for the responsibility of the person processing their likeness.

AI serves as a tool here, not an excuse that absolves personal agency. Selecting an image, assigning it a new context, and distributing the result are separate steps. Differentiating them matters, because halting the creation of new fakes is not the same as curbing the spread of existing ones. An incident response focused purely on identifying the technology misses the operational question of what to do with media already circulating.

Reporting an incident must also be distinguished from republishing the offending material. Outrage in a comment section does not alter the fact that another copy is distributed. Good intentions are an insufficient defense: what matters is whether sharing the file is strictly necessary for an investigation. An advisory or warning can outline the issue without reposting the image. Minimizing exposed detail is not trivializing the incident; it prevents creating another distribution vector.

Outcomes vs. Potential Impact

This case provides no grounds to tally winners and losers. The official statement alone does not demonstrate that the individual regained control over their likeness or that distribution was stopped. However, one can outline the potential costs and benefits of an organized response. The victim benefits from a process that allows reporting without triggering further public exposure. That is an operational goal, not a verified final outcome of the Kraków matter.

An organization, school, or community moderator benefits from clear responsibilities. Intake, triage, and resolution updates should not fall on random individuals. The trade-off is the operational time required to handle reports responsibly. Swift handling is only valuable if it mitigates the issue, not if it simply rushes to close the ticket.

A viewer who chooses not to repost gives up satisfying public curiosity, but stops an additional link in the chain. It is an unglamorous role that offers no public proof of advocacy. This is why actions should be judged by their effects rather than their visibility. Refraining from publication often protects the individual far better than vocal condemnation accompanied by a shared image.

Actionable Lessons for Organizations

An effective policy does not require the reporting party to identify the AI tool used or prove a legal violation. It should simply allow them to report the problem and provide essential details for review. Spotting deepfakes should never be a prerequisite for filing a complaint. Organizations need clear escalation paths, and users need a simple intake channel that does not encourage public circulation.

• Designate a restricted intake channel and assign a specific response lead. Restrict internal access to the report rather than sharing it across entire teams.
• Ensure intake confirmations distinguish between acknowledging an inquiry and resolving it. Promising complete eradication across the web requires verified capabilities, not optimistic assumptions.
• Advise users to stop forwarding the material, log the timestamp and URL where it was encountered, and report it to platform moderators or relevant authorities.
• If the necessary reporting details are unclear, consult the intake team rather than circulating the file further to gather opinions.

Scenario: Reporting Within an Internal Group

Consider a scenario where an altered image of an employee appears in an internal workspace. This hypothetical applies the core principles without recreating the Kraków incident. The first decision involves alerting others: do you display the image so everyone understands the issue, or describe the breach without sharing the file? The safer approach is an advisory against redistribution paired with a link to the reporting channel. Staff members do not need to view the material to understand internal policy.

The next challenge is documentation. The administrator needs sufficient data to locate the file, but distributing copies to all executives is unnecessary. The URL and discovery timestamp can be logged, with further evidence coordinated directly with the incident response lead. The objective is mitigation, not maintaining a widely accessible internal archive of the deepfake. Removing the post from public channels and logging evidence must be treated as separate tasks.

Finally, communication with the depicted individual must be handled carefully. Updates should detail concrete actions taken and provide a clear contact point, without overpromising that all copies have been eliminated. Rushing to mark an issue resolved may ease internal workflows, but it does not mean the material has vanished. Mitigate exposure while preserving evidence, and keep affected parties informed without overstating the outcome.

Unknown Factors

The ultimate outcome of the police referral and the total audience reach remain unknown. UODO, June 24, 2026. These gaps should not be filled with speculation about motives or personal impacts. The core takeaways remain evident: technical access does not grant permission, lack of authenticity does not negate harm, and opening an inquiry is not a final ruling. Procedures and individual sharing habits should be built on these distinctions. Further analyses are covered in our series on AI real-world cases.

Guidance for Witnesses

Anyone who encounters an unauthorized altered image needs a clear protocol: do not redistribute it, document where it appeared, and submit it through the proper reporting channel. Seeking assistance should not require detailing the event publicly. Procedures should also accommodate third-party reports when the affected individual prefers not to contact administrators directly.

Google · Your sourcesQuantslant higher in your Google - one clickAdd to preferred sources →

Sources

• UODO: Kraków minor deepfake case - June 24, 2026.
• UODO: Deepfakes and illegal online content - September 22, 2026.