Scatter Lab has received the questionable honour of being the first South Korean AI tech startup to be fined for indiscriminately sourcing data from other platforms to train what turned out to be a racist dating chatbot. The harvested personal data included info on 200,000 children under the age of 14; to make thing worse, the training data – which included names and other personal information – was later uploaded to GitHub.
The Personal Information Protection Commission (PIPC) ordered the Seoul-based company to pay ₩103.3m ($92,900) in penalties for its flagrant disregard for people’s privacy.
Scatter Lab is accused of collecting 600,000 people’s 9.4 billion conversations on chat platform KakaoTalk from its Science of Love and Text At apps to train chatbot Lee Luda without users’ permission, news agency Yonhap reported. The virtual agent was designed for people who’d failed to find companionship with real people.
Science of Love had been a smash hit for Scatter Lab, downloaded a total of 7.5 million times by 2020 according to The Register. The app’s coup de grace was a feature in which users could send their conversations with their romantic interests on KakaoTalk to Science of Love and get an analysis in return about just how engaged the other person really was in the conversation.
Scatter Lab’s troubles started when it decided to expand its offerings to include Lee Luda. Users quickly grew weary after the chatbot, which poised as a 20-something college student, assaulted them with homophobic and racist remarks as well as offensive statements about people with disabilities. As a result, the online reviews on Google plummeted.
But that wasn’t the end of it. Not only did Scatter Lab harvest personal data from the conversations without users’ consent – or from guardians and parents in the case of the children – but the training data was also uploaded to GitHub.
It is against this background that the PIPC has now whacked Scatter Lab across the fingers. “This is the first case of restricting the indiscriminate processing of personal information by AI technology companies, restricting the use of personal information collected for specific services by companies for other services without the user’s explicit consent, through which AI development and services are provided,” the regulator said in a statement, translated with Google Translate.
Scatter Lab has seemingly accepted the fines and its own wrongdoing, pledging to do better in the future.
“We feel a heavy sense of social responsibility as an AI tech company over the necessity to engage in proper personal information processing in the course of developing related technologies and services,” the company said in a statement, adding that it would “not only actively implement the corrective actions put forth by the PIPC but also work to comply with the law and industry guidelines related to personal information processing.”