CourseNatural Language Processing: The Complete Course · Module 10: LLMs for NLP, and Responsible NLP · part 71 of 73
Part 71 · Module 10: LLMs for NLP, and Responsible NLP

Incident: A chatbot that repeated private chats

30 min read·14 Oct 2026
Module 10 · Incident file · About 30 minutes

What happened

Dec 2020 launched on Messenger training data private chats from an app within weeks hateful remarks, personal details Jan 2021 suspended; a fine came later Scatter Lab's Lee Luda chatbot, South Korea
The reported case in order: a launch (blue), training on private chats (amber), the harm within weeks (red), then suspension and a later fine (grey).

Lee Luda was a chatbot that the South Korean company Scatter Lab launched on Facebook Messenger in December 2020. It was built from billions of private KakaoTalk messages, collected through the company's dating-advice app, where people had shared their chats.

Within weeks it made hateful remarks about minorities, and users found it repeating personal details. Scatter Lab suspended it in January 2021. South Korea's Personal Information Protection Commission later fined the company over how the data was used.

Two harms came from one cause. Users of a public chatbot met prejudice and other people's details, and private conversations had been turned into a product's training data, which the regulator later fined. Both follow from training a model on private messages without removing what must never come back out.

The rest of this part opens with an account

Signing up is free and takes one email. Nothing to pay, and nothing renews.

Start reading free