Find the Right GRC Solution

Search and compare GRC technology built for the problems you’re trying to solve

Belgian Privacy Regulator Orders AI Chatbot Company to Comply With GDPR Over Sensitive User Data

Belgian Privacy Regulator Orders AI Chatbot Company to Comply With GDPR Over Sensitive User Data

By
🔒
Cancel anytime
Full archive access
Custom alerts
Key Takeaways
  • Sensitive Data Processing Violations: Belgium's Data Protection Authority found that an AI chatbot application lacked the explicit consent required to use sensitive conversational data for model training.
  • Legitimate Interests Accepted for Ordinary Data: The regulator determined that the company could rely on legitimate interests to process ordinary conversational data for AI model improvement under the specific circumstances examined.
  • Transparency Deficiencies Identified: The company had not adequately explained how ordinary and sensitive conversational data were processed. It updated its privacy policy concerning ordinary data during the proceedings.
  • Four-Month Compliance Deadline: The company must correct its processing of sensitive information and improve its disclosures or face penalty payments totaling up to €2,000 per day if both requirements remain unmet.
Deep Dive

Belgium's Data Protection Authority has given the operator of an AI chatbot application four months to correct its handling of sensitive user information after finding that the company used personal data to train its models without obtaining the consent required under European privacy law.

The decision concerns an application that allows users to create personalized chatbots. The regulator found that the company could lawfully use ordinary conversational data to improve its AI models, but had failed to establish the necessary legal grounds for processing sensitive information disclosed during those conversations.

This proved decisive in a case that began with a complaint alleging the company had processed users' personal information without a valid legal basis to train and deploy its chatbots. Belgium's Data Protection Authority (APD) examined the allegations under the General Data Protection Regulation (GDPR), separating ordinary conversational data from information subject to the regulation's stricter protections.

The authority's Litigation Chamber accepted the company's reliance on legitimate interests for processing ordinary conversational data. Improving the consistency of AI-generated responses and the relevance of chatbot interactions served a legitimate purpose, it found, and the processing was necessary to achieve those objectives.

The chamber also considered what users could reasonably expect when interacting with the application. It determined that users could anticipate their ordinary conversational data being used to improve the service, particularly given the safeguards the company had established to protect their rights and freedoms.

That reasoning did not extend to sensitive personal information. Under the GDPR, certain categories of data receive additional protection because of the nature of the information they reveal. These include details concerning an individual's health, religious beliefs or sexual orientation, including information from which such characteristics can be inferred.

Processing these categories of information is generally prohibited unless a specific legal exception applies. Explicit consent is one such exception, but the regulator found that the company had not obtained it before using sensitive conversational data to train its AI model.

The authority consequently rejected the company's processing of that information, even as it accepted the use of ordinary conversational data for the same general purpose.

The investigation also identified shortcomings in the company's explanations to users about how their information was handled. According to the regulator, the application had not provided sufficient transparency concerning the processing of ordinary and sensitive conversational data used to operate and train its chatbots.

During the proceedings, the company updated its privacy policy regarding ordinary conversational data. The regulator acknowledged those changes but concluded that additional measures were necessary to address the processing of sensitive information.

The Litigation Chamber issued a reprimand and ordered the company to bring two aspects of its operations into compliance. The first concerns its use of sensitive conversational data to optimize its AI model. The second concerns the information provided to users about that processing.

The company has four months to satisfy both requirements. Failure to comply could result in penalty payments of €1,000 per day for each outstanding obligation, amounting to €2,000 daily if neither requirement is met.

The regulator chose the compliance order in light of measures the company had already undertaken, allowing it to continue the corrective work begun during the proceedings rather than imposing an immediate administrative fine. The parties have 30 days to appeal the decision.

The ruling is a relatively narrow one, but it extends to a question confronting companies that train AI systems on conversations with their users. Information collected through a single interaction may fall under different provisions of the GDPR, and a lawful basis for processing one category does not necessarily permit the processing of another. Belgium's authority did not conclude that chatbot operators must obtain consent before using all conversational data for model improvement. Its findings instead turned on the nature of the information involved, the purpose of the processing and the safeguards available to users.

 The GRC Report is your premier destination for the latest in governance, risk, and compliance news. As your reliable source for comprehensive coverage, we ensure you stay informed and ready to navigate the dynamic landscape of GRC. Beyond being a news source, the GRC Report represents a thriving community of professionals who, like you, are dedicated to GRC excellence. Explore our insightful articles and breaking news, and actively participate in the conversation to enhance your GRC journey.

Oops! Something went wrong