Major privacy laws - including GDPR - could be downgraded to try and boost AI growth and cut red tape (www.techradar.com)

European officials are reportedly preparing a “digital omnibus” proposal due November 19, 2025, that could loosen how some European privacy rules—potentially including parts of the GDPR—apply to AI development. Drafts seen by Politico indicate regulators may reclassify pseudonymized data (records stripped of direct identifiers) so it’s not always protected as personal data, and could permit processing of sensitive categories (political opinions, religion, health) for model training. The package may also broaden legal bases for online tracking beyond explicit consent, while claiming changes would be “targeted” so core GDPR principles remain intact. The European Commission has made no public announcement yet, and the draft has sparked pushback from several member states and privacy pioneers including GDPR architect Jan Philipp Albrecht. For the AI/ML community this is consequential: easing data restrictions would unlock larger, richer training sets—potentially narrowing the competitiveness gap with the U.S. and China and reducing regulatory roadblocks that have delayed deployments by Big Tech. But it also raises major compliance and technical implications: stricter requirements for pseudonymization/anonymization standards, updated lawful-basis and DPIA practices, renewed focus on provenance and model auditing, and heightened political/legal risk if privacy protections are perceived as weakened. Developers, privacy teams and regulators will need to reconcile innovation goals with safeguards that prevent re-identification and misuse of sensitive attributes.