Antoon Bever

Dutch intelligence agencies got caught training AI on citizen data they weren't supposed to have. the CTIVD, the Netherlands' official intelligence oversight body, published a report finding that AIVD and MIVD staff accessed and retained massive bulk datasets in ways that violate Dutch law. names. phone numbers. location data. social media. communication content. millions of entries. some from government sources. some commercially purchased. some stolen by criminals and bought on the dark web. and then Bits of Freedom, the Dutch digital rights organization, flagged something buried in the report: the agencies appear to be training their own AI models on this data. this is not the first time. in 2020, the same watchdog found the same agencies had retained citizen data far beyond legal limits. after complaints, they were ordered to delete it. six years later: same report. same finding. same agencies. but now with AI in the loop. the data that was illegally retained is now reportedly being used to train models that will make future surveillance faster, more accurate, and more autonomous. you cannot opt out. Unlike commercial AI companies, there is no settings menu. no privacy center. no right to delete that you can actually exercise. Bits of Freedom put it directly: "They seem to be buying data from criminal data breaches. aren't they supposed to be protecting us from those?"