Clarifai deleted 3 million OkCupid photos and the face recognition models trained on them

Clarifai reported that it deleted approximately 3 million photographs of OkCupid users, as well as face recognition models trained on these images. The story dates back to 2014: at that time, the data was transferred to the AI company without the knowledge of the users themselves and in violation of the dating service's privacy policy.

How the Story Emerged

Under the terms of a settlement with the U.S. Federal Trade Commission, the dating platform OkCupid transferred a large batch of user photos to Clarifai for computer vision tasks. The key problem was not only the data transfer itself, but the fact that people did not know about it and did not give separate consent to use their images for training algorithms. Moreover, such an exchange, as indicated in the case materials, contradicted OkCupid's own privacy policy.

The story is particularly sensitive because of the type of data involved. This was not just pictures for a general dataset, but photographs of faces that could potentially be used to train recognition systems. When such data is transferred without clear notification and without explicit consent, the dispute quickly goes beyond the usual conflict over user terms and becomes a question of biometric privacy. This is why such cases are increasingly being considered separately from ordinary disputes about user data.

What Clarifai Deleted

Clarifai, an American company from Delaware working in the field of computer vision and face recognition, confirmed that it erased approximately 3 million images received from OkCupid. Along with the original files, the models trained on these photographs were also deleted. This is an important point: if you remove only the dataset but leave the derivative models in operation, the problem of data origin does not disappear. The regulatory significance in this story is precisely in the destruction of both layers.

• Approximately 3 million user photographs
• The data transfer occurred back in 2014
• Users were not notified in advance of this use
• Not only the images were deleted, but also the models trained on them

Clarifai itself was not accused by the FTC of violations. This separates its role from that of OkCupid and Match Group, which appeared in the settlement as parties responsible for user data practices. But the public confirmation of deletion shows that the company was nonetheless forced to respond to the old data transfer and close the issue not only formally but also technically. This is a rare case where the issue is closed with explicit deletion of derivative artifacts.

Why This Matters Now

The FTC announced a settlement with OkCupid and Match Group in late March 2026, and there were no financial penalties in it. This outcome appears lenient given the scale of the issue and the sensitivity of facial photographs. The regulator focused not on monetary punishment, but on the fact that the service violated its own promises to users and allowed data to be transferred for purposes about which the audience was not properly informed.

The absence of a fine does not make this case insignificant. On the contrary, it demonstrates how long the consequences of decisions made in the early years of the AI market can persist, when companies were more eager to collect and reuse data than to explain it in detail to people. Twelve years after the photos were transferred, the question still came back — already in the context of the regulator, privacy, and deletion of models, not just the original files.

For the industry, this is almost a textbook example of deferred regulatory risk. For Match Group and other platforms, it is also a reminder that an old integration with an external AI contractor can suddenly become a reputational problem.

A user who uploaded a photo to a dating app probably did not expect that these images would help train a face recognition system. It is precisely this gap between audience expectations and actual data use that has become the main source of risk today. Especially when it comes to faces rather than neutral images of objects.

What This Means

For AI companies, this is a signal to more carefully check the provenance of datasets, especially if they involve faces and other sensitive information. For digital platforms, it is a reminder that old data transfers can resurface years later and lead to requirements to delete not only the original files but also the results of training. And for the market as a whole, it is another step toward the rule: contentious data cannot simply be "archived," it must be removed from models completely.