Lee Luda AI chatbot spouts offensive responses

South Korean chatbot Lee Luda's offensive behaviour and abuse of users' privacy led to it being shut down three weeks after its lauch.

Launched in December 2020, Lee Luda was a Facebook Messenger-based chatbot developed by south Korean web firm Scatter Lab to mimic a 20-year old female university student and potential virtual starlet. Lee Luda was trained using over 10 billion KakaoTalk messages from two other apps developed by the same company. 

The initial reaction to Lee Luda was broadly favourable, with 750,000+ people, mostly teenagers and young adults, subscribing to its Facebook Messenger account. They were mostly drawn by its informality and ability to build 'intimacy' over time, according to the Dankook Herald.

However, the bot quickly started spraying users with offensive, discriminatory, and homophobic comments about women, lesbians, black, disabled and trans people, and was used to sexually harass others.

Lee Luda was also accused of disclosing the personal data of people whose data was included in the training of its model. In April 2021, Korea’s data protection watchdog the Personal Information Protection Commission (PPIC) fined Scatter Lab 103.3 million won (USD 93,000) for not obtaining proper user permissions.

The fiasco was seen to prompt debate about AI ethics in South Korea, culminating in the Korea AI Ethics Association (KAIEA) calling for the immediate suspension of the service and a coalition of civil society organisations denouncing the government's promotion of the AI industry at what it saw as the expense of digital rights, and calling for a more stringent regulatory framework.

Regulatory, legal 👩🏼‍⚖️

• Korea Personal Information Protection Commission (2021). 개인정보위,‘이루다’개발사 ㈜스캐터랩에 과징금·과태료 등 제재 처분

News, commentary, analysis 🗞️

• https://koreajoongangdaily.joins.com/2021/01/11/business/industry/AI-AI-chatbot-Lee-Luda/20210111175200496.html

• http://dkherald.dankook.ac.kr/news/articleView.html?idxno=12178

• https://en.yna.co.kr/view/AEN20210111009800325

• https://www.koreatimes.co.kr/www/tech/2021/01/133_302390.html

• http://www.koreaherald.com/view.php?ud=20210111001051

• http://www.koreaherald.com/view.php?ud=20210122000620

• http://koreabizwire.com/ai-chatbot-creator-to-suspend-service-amid-controversy-over-vulgar-language/179135

• https://www.vice.com/amp/en/article/akd4g5/ai-chatbot-shut-down-after-learning-to-talk-like-a-racist-asshole

• https://www.independent.co.uk/news/world/asia/facebook-hate-speech-ai-chatbot-south-korea-b1787181.html

• https://thenextweb.com/news/chatbot-shut-down-after-saying-it-really-hates-lesbians-and-using-racist-slurs

• https://www.inputmag.com/culture/south-korean-chatbot-lee-luda-killed-off-for-spewing-hate

• https://www.scmp.com/news/asia/east-asia/article/3117723/south-korean-firm-takes-down-lgbt-hating-chatbot-over-hate

• https://thediplomat.com/2021/01/chatbot-gone-awry-starts-conversations-about-ai-ethics-in-south-korea/

• https://www.dw.com/en/deepfakes-rattle-south-koreas-tech-culture/a-56310213

• https://slate.com/technology/2021/04/scatterlab-lee-luda-chatbot-kakaotalk-ai-privacy.html

• https://www.theregister.com/2021/04/29/scatter_lab_fined_for_lewd_chatbot/