Secret EU plans to allow Big Tech to train AI with our personal data (www.brusselstimes.com)

Leaked reports indicate EU insiders are preparing proposals that would open the door for large tech firms to use personal data to train AI models — a move described as “secret” in coverage. If enacted, the change would mark a significant shift from the EU’s strict data-protection posture under GDPR by carving out or relaxing rules that currently limit how identifiable or sensitive personal data may be processed for model training. The story matters because it could reshape incentives and legal risk across the AI ecosystem: Big Tech firms with vast data troves would gain a competitive edge, while users, privacy advocates and smaller AI developers would face new exposure and uncertainty. Technically, training on raw personal data increases risks of model memorization and unintended disclosure of sensitive attributes or individual records, heightening re-identification threats. Mitigations exist — differential privacy, federated learning, secure multi-party computation, synthetic-data generation, strict access controls and independent model audits — but they vary in effectiveness and cost. The debate will hinge on whether the EU ties any new permissions to strong technical safeguards, oversight mechanisms, enforceable data-subject rights (like erasure and consent), and compensation or governance models (data trusts, transparency logs). The outcome will influence research practices, market competition, and the practical meaning of privacy protections across Europe.