Google scientist warns EU data plan
- Google’s Sergei Vassilvitskii told EU regulators on May 5 that a plan forcing Google to share search data with rivals could expose users. - The flashpoint is the EU’s anonymisation idea: Google says its red team could reverse it in under two hours. - That matters because Brussels wants AI chatbots and rival search engines to get Google search signals under the DMA.
Search data is the raw exhaust of how people think — what they want, what they fear, what they almost typed and then refined. That is why the EU’s latest fight with Google is bigger than a normal antitrust skirmish. Brussels wants Google to share parts of that data with rival search engines and AI products under the Digital Markets Act. Google’s pushback this week was blunt: its own privacy scientists say the EU’s anonymisation plan may not really anonymise much at all. (digital-markets-act.ec.europa.eu) ### What did Google actually warn? On May 5, Sergei Vassilvitskii — a distinguished scientist at Google focused on differential privacy — told EU antitrust regulators that the proposed data-shari(digital-markets-act.ec.europa.eu)hatbots with search features. Google is basically saying the competition remedy collides with the privacy problem. (digital-markets-act.ec.europa.eu) ### What is the EU trying to force Google to share? The Commission’s April 16 proposal says Google should give third-party search engines access to search data on fair, reasonable, and non-discri(digital-markets-act.ec.europa.eu)t is simple: help rivals improve their products enough to challenge Google Search. (digital-markets-act.ec.europa.eu) ### Why is anonymisation the whole fight? Because search logs are weirdly identifying even when obvious names are stripped out. A search trail can still reveal a person through patterns — locatio(digital-markets-act.ec.europa.eu)ons intact. Google’s argument is that modern AI makes that reconstruction easier, not harder. (thenextweb.com) ### What was the “two hours” claim? That came from Google’s internal red-team testing, as described in coverage of the meeting. Vassilvitskii’s warning was that the Commission’s proposed anonymisation method could be reversed in les(thenextweb.com)aim holds up, the remedy is not just risky in theory — it is breakable on a normal workday. (thenextweb.com) ### Why does OpenAI keep showing up in this story? Because the Commission’s proposal is not only about classic search engines. It explicitly contemplates access for AI chatbots with search functionality. That means the fight is real(thenextweb.com)best behavioral data locked inside the company. (digital-markets-act.ec.europa.eu) ### Is this just Google protecting its moat? Partly, yes. Google has every incentive to resist handing over the feedback loops that help tune search quality. But that does not make the privacy co(digital-markets-act.ec.europa.eu) makes data remedies so much harder than forcing a company to unbundle an app or change a default. (digital-markets-act.ec.europa.eu) ### What happens next? The Commission has been consulting on the proposed measures and is expected to decide on final steps later in the summer. So the near-term question is not whether Europe wa(digital-markets-act.ec.europa.eu) ### Bottom line? This is a test of whether antitrust can reach into data flows without breaking privacy on the way. If Brussels gets it right, rivals get a real shot. If it gets it wrong, the price of more competition could be the search histories of ordinary people.
