Bias, Discrimination and Data Protection in the Age of GPAI
summary
A panel documents that general‑purpose AI models trained on massive scraped and first‑party corpora — for example Meta training on Facebook/Instagram first‑party data — produce layered biases that pit accuracy against fairness, with concrete failures like Gemini's black Vikings controversy showing problems aren’t solved by more data. Regulators must untangle GDPR Article 22 and the AI Act, consider social license for data reuse and data commons, and treat GPAI as a systemic risk that can amount to an instruction to discriminate.
AI-written summary · about
chapters (6)
- 0:00Panel on GPAI bias
- 19:38Discrimination law basics
- 36:44Statistical data contradictions
- 54:19Reconceptualizing discrimination law
- 1:00:01Digital self-determination
- 1:09:12Q&A
related talks
01h 19mAligning Data Privacy and AI Governance
01h 14mLatest Developments in AI and Data Protection Legal Uncertainty Despite
28mEthical AI?!
53m10 Key Challenges for AI within the EU Data Protection Framework
11mAddressing Privacy, Bias, and Fairness in AI Systems
01h 11mTransforming GDPR into a Risk Based Harm Tool Alongside Specific AI Regulation...
01h 08mTHE GOVERNANCE OF AI: CONVERGENCE OR DIVERGENCE?
01h 11mBack to Basics How Data Protection Principles Apply to AI Systems