Clarifai Deletes 3 Million OkCupid Photos Used for Facial Recognition Training After FTC Settlement

In a rare case of delayed accountability, AI startup Clarifai has deleted 3 million OkCupid profile photos it used to train facial recognition models in 2014—12 years after the data was collected. The deletion follows a settlement reached last month between the Federal Trade Commission (FTC) and Match Group, OkCupid’s parent company.

The Delaware-based Clarifai certified the data deletion to the FTC on April 7 and confirmed to U.S. Representative Lori Trahan (D-MA) that it had also removed any models trained on the OkCupid data. The company stated it had not shared the data with third parties.

FTC Investigation and Privacy Violations

The FTC opened its investigation in 2019 after The New York Times reported that Clarifai had built a training database using OkCupid profile photos, directly violating the dating site’s privacy policy. Court documents reviewed by Reuters show that Clarifai requested the data from OkCupid executives in 2014 and received it.

Clarifai used the photos to develop a facial recognition service capable of identifying a person’s age, gender, and race. The company’s founder, Matthew Zeiler, emailed OkCupid co-founder Maxwell Krohn in 2014, stating,

"We're collecting data now and just realized that OkCupid must have a HUGE amount of awesome data for this."

Ethical Concerns and Past Practices

In 2019, Zeiler downplayed privacy concerns, telling The New York Times,

"There has to be some level of trust with tech companies like Clarifai to put powerful technology to good use, and get comfortable with that."

Notably, some of OkCupid’s founders were also investors in Clarifai, raising further questions about conflicts of interest.

FTC Settlement and Its Limitations

As part of the settlement, the FTC imposed a permanent ban on OkCupid from misrepresenting its data collection and privacy controls. However, TechCrunch highlighted the irony of this penalty, noting that FTC rules already prohibit such misrepresentations. The settlement did not include financial penalties or additional restrictions.

The case underscores ongoing concerns about data privacy, consent, and the ethical use of personal information by AI companies, even a decade after the initial violation.