Michael Veale

@[email protected]
6K Followers
1.3K Following
2.4K Posts

Professor of Technology Law and Policy at University College London (#UCL), Faculty of Laws.
Not resigned to today's technological power structures (yet). Researching at the intersection of emerging technologies, law and policy; data protection; machine learning; PETs and cryptographic infrastructures; platform and infrastructural regulation. 🏳️‍🌈

administrating a small exoplanet in the fediverse .

websitehttps://michae.lv
matrixmv:matrix.michae.lv
encrypted emailhttps://files.michae.lv/encryption.txt
bsky@michae.lv
Michael Veale5h ago
OpenAI gets further and further down the platform speed run. If you communicate content to users that you do not, or do not wish to, control, then liability claims come your way sooner rather than later. https://www.theglobeandmail.com/canada/article-sam-altman-openai-tumbler-ridge-apology/
OpenAI’s Altman ‘deeply sorry’ company didn’t flag Tumbler Ridge shooter’s messages to police

Letter to community is ‘necessary’ but ‘grossly insufficient,’ B.C. Premier Eby says

The Globe and Mail
Show threadMichael Veale6h ago
special call out to @tmsqrll of this here server
Michael Veale13h ago
Great work from Foxglove — UK government sheepishly forced to increase estimated carbon emissions from data centre usage by... 100 times https://www.ft.com/content/0c8bc0a9-63e5-4739-a91e-f07189b45f20?syn-25a6b1a6=1
AI data centre emissions vastly underestimated, UK admits

New projections raise forecasts of climate impact by up to 136 times

Financial Times
Michael Veale1d ago
Short and strong piece on why user friendly edtech, AI tutors and the like, endangers learning by removing the productive friction. Constrained AI might be seen as a solution, but is that then really better than education without AI? a roundtable chat in a room? https://link.springer.com/article/10.1007/s13347-026-01093-6
Why the User-Friendliness of AI Undermines Education. Commentary on Wieczorek 2025 - Philosophy & Technology

In his article “Why AI Won’t Democratize Education”, Wieczorek (2025) challenges the techno-optimistic narrative surrounding intelligent

SpringerLink
Michael Veale1d ago
a reminder of a powerful book review from 2019 https://newrepublic.com/article/154236/sameness-cass-sunstein
The Sameness of Cass Sunstein

His books keep pushing the same technocratic fixes. But today’s most pressing questions cannot be depoliticized.

The New Republic
Michael Veale2d ago
AI overviews pulling in through RAGs old gov pages that are there for archiving and poserity, wreaking havoc. DoE advises to “design w/ the expectation that much of what we publish will be [..] atomized, summarized or reinterpreted by systems we don't control”
https://www.theregister.com/2026/04/23/stale_govuk_pages_are_feeding/
Stale gov.uk pages are feeding AI overviews old data and Brits are believing it

: Whitehall content teams play whack-a-mole with zombie pages as Google hoovers up the lot

The Register
Michael Veale2d ago

More TV specific US privacy law, this time from Kentucky, as ‘smart television’ data is determined as sensitive by HB 692 https://apps.legislature.ky.gov/recorddocuments/bill/26RS/hb692/bill.pdf

vibes of the 1988 Video Privacy Protection Act 1988, specifically protecting eg Blockbuster history https://www.congress.gov/bill/100th-congress/senate-bill/2361

Michael Veale2d ago
Stable Firefox identifier (the order of a set of IndexedDB databases) allowed cross-site linkage within a single browsing session, private window or not — also in the Tor Browser https://fingerprint.com/blog/firefox-tor-indexeddb-privacy-vulnerability/
We Found a Stable Firefox Identifier Linking All Your Private Tor Identities

We discovered a privacy vulnerability in Firefox Private Browsing and Tor Browser that allows websites to fingerprint and track users across origins using IndexedDB database ordering, even after closing all private windows.

Fingerprint
Michael Veale2d ago
Republicans introduce ‘Secure Data Act’, an attempt to copy some of the weakest US state privacy laws and apply them across the country to eg pre-empt the strong ones like California or Colorado. No private right of action, 45 day ‘curing’ period for any violations. https://www.cnbc.com/2026/04/22/data-privacy-bill-congress-states.html
Michael Veale2d ago

Interesting recent blog on reverse engineering Google SynthID (images) and adding fake watermarks in that Google detects. Also on the stupidity of making Gemini the detector, as it hallucinates its own detection capabilities… https://hackerfactor.com/blog/index.php?/archives/1092-Reversing-SynthID.html

repo: https://github.com/aloshdenny/reverse-SynthID