GDPR Compliance for AI Systems: Decoding the Belgian DPA’s Guidance

Introduction:?

The intersection of artificial intelligence (AI) and data protection law has become increasingly important in today’s digital landscape. As AI technologies progress, they introduce unique challenges that call for a deep understanding of legal frameworks surrounding Data Protection and AI.???

To address the issue, the Belgian Data Protection Authority (DPA) has published guidelines titled 'Artificial Intelligence Systems and the GDPR – A Data Protection Perspective,' explaining the GDPR requirements specifically applicable to the development and deployment of AI systems. The GDPR offers a strong framework for safeguarding personal data within the EU. Simultaneously, the EU AI Act introduces new provisions that specifically tackle the challenges related to high-risk AI systems. The guidelines issued by DPA is aim to ensure that AI systems align with data protection principles so that they operate ethically.??

What is an AI System??

The guidance provides a detailed definition of an AI system. According to the EU AI Act, an AI system is a computer system specifically designed to analyse data, identify patterns, and apply that knowledge to make informed decisions or predictions. In some instances, AI systems can learn from data and adapt over time, improving their performance, identifying complex patterns across various data sets, and making more accurate or nuanced decisions.?

A common example of AI systems in daily life is virtual voice assistants, like Siri or Alexa. These systems process voice commands, identify speech patterns, and provide relevant responses or actions. Over time, as users interact with them, the AI learns from these interactions, becoming better at understanding accents, preferences, and more complex commands.?

Key GDPR Principles for Ethical AI Deployment:?

The GDPR outlines essential principles for lawful processing of personal data within AI systems, which are also reinforced by the EU AI Act. Key principles include:?

• Lawfulness: AI systems must adhere to legal processing standards, with specific prohibitions on high-risk applications like social rating systems and real-time facial recognition in public spaces.?

• Fairness: The EU AI Act emphasizes fair processing, focusing on reducing bias and discrimination in AI deployment.?

• Transparency: Users should be informed when engaging with AI systems. High-risk applications require detailed instructions on their capabilities and limitations.?

• Purpose Limitation and Data Minimization: Data must be collected for legitimate, specified purposes, ensuring no excessive data is gathered.?

• Data Accuracy: AI systems must use high-quality, up-to-date data to prevent discrimination.?

• Automated Decision-Making: Individuals can contest automated decisions, with human oversight mandated for high-risk AI.?

• Accountability: Organizations must document AI systems and implement oversight measures to ensure compliance with GDPR principles.?

Simplifying Compliance with GDPR and AI Act:?

This DPA guidelines bridge the gap between legal obligations and AI system development. Key principles include ensuring lawful data processing by assessing legal bases, promoting fairness by mitigating bias, and maintaining transparency about data usage. It provides that organizations must also adhere to purpose limitation and data minimization principles, ensuring data accuracy and security. Furthermore, the AI Act reinforces the rights data subjects have under the GDPR by requiring clear explanations about how their data are used. Additionally, demonstrating accountability through thorough documentation and conducting Fundamental Rights Impact Assessments (FRIAs) is essential to mitigate potential risks associated with AI systems.?

Conclusion:?

The Belgian DPA’s guidelines ensure AI systems comply with GDPR and the AI Act by promoting lawful, fair, and transparent data processing. Organizations must ensure accountability through clear documentation and risk assessments, aligning with both the GDPR and the EU AI Act.?

If you’re an organization dealing with copious amounts of data, do visit www.tsaaro.com .??

Read our blog on- Conducting DPIAs for AI Systems: Navigating Ethics and Data Privacy ?

?

News of the Week?

?

1. Virginia Prosecutor Sues Georgetown University Over Data Breach?

?Mary Margaret Cleary, a deputy commonwealth attorney for Culpeper County and Georgetown alumna, filed a class action lawsuit against Georgetown University following a data breach that exposed personal information of students, including Social Security numbers, tax ID numbers, and employee payroll data. The breach occurred due to an internal error, allowing 29 students to access the information for 24 hours. Cleary alleges increased risks of identity theft and financial fraud, seeking damages for the potential harm caused.?

https://www.reuters.com/legal/litigation/virginia-prosecutor-sues-alma-mater-georgetown-over-data-breach-2024-10-18/ ??

?

?

2. X Updates Privacy Policy to Enable AI Training on User Data?

?X (formerly Twitter) has updated its privacy policy, effective November 15, allowing third-party collaborators to use user data to train AI models unless users opt out. Notifications about the changes have been sent, with the platform aiming to generate revenue by licensing this data. X is currently excluding EU users due to stricter data laws but hasn’t clarified if users elsewhere can opt out later. The policy introduces a $15,000 penalty for accessing over a million posts daily, targeting misuse of large-scale tweet extraction.?

https://www.moneycontrol.com/technology/elon-musks-x-updates-privacy-policy-to-train-ai-models-on-your-posts-article-12845201.html ??

?

?

3. U.S. Department of Labor Releases AI Guidelines for Workplace?

?On Oct. 16, the U.S. Department of Labor issued detailed AI principles, expanding on President Biden’s 2023 executive order. The guidelines emphasize ethical AI development, transparency, data privacy, and collective bargaining rights. They also promote responsible AI use, warning of bias and job displacement risks, while encouraging worker protection, training, and meaningful human oversight to ensure AI enhances job quality, particularly for underserved communities.?

https://www.dol.gov/newsroom/releases/osec/osec20241016 ??

?

?

4. Internet Archive Faces Third Cyberattack?

The Internet Archive suffered its third security breach on October 20, 2024, as hackers exploited unrotated Zendesk API tokens. Despite prior warnings, the organization failed to secure these tokens, allowing access to sensitive support data, including personal identification documents. This breach follows two previous attacks earlier in the month, exposing a critical flaw in the Archive's security practices and token management.?

https://www.forbes.com/sites/larsdaniel/2024/10/20/internet-archive-breached-again-third-cyber-attack-in-october-2024/ ??

?

?

?5. Bundestag Approves Germany’s New Regulation to Reduce Cookie Banner Overload?

The German government has introduced a regulation aimed at reducing the constant flood of cookie banners by allowing users to give consent through recognized services. It introduces a framework for consent management services, allowing users to centrally manage consent for cookies and similar technologies. Companies can opt to integrate these services instead of using traditional cookie banners, though integration remains voluntary.??

https://www.heise.de/en/news/Data-protection-Bundestag-approves-regulation-against-cookie-banner-flood-9985453.html ??

?