michael veale
LightNews LightNews
banner
michae.lv
michael veale
@michae.lv
8.4K followers 1.8K following 1.4K posts
prof @laws.ucl.ac.uk, technology, law, policy, society, whimsical latvian top level domain names. michae.lv and fediverse https://someone.elses.computer/@mikarv 🏳️‍🌈
Posts Replies Media Videos
michae.lv
On AI’s ‘mediocrity trap’ — experiments indicates that while AI helps the less skilled make something passable, the highly skilled don’t use it to produce something better than they could have; they produce something ok, but lose motivation to make it great. www.jin-li.org/uploads/1/1/...
Page 10-11 of the linked PDF
November 26, 2025 at 9:18 PM
michae.lv
includes useful visuals!
November 18, 2025 at 2:00 PM
michae.lv
New from @jpquintais.bsky.social & me: a critical and wide analysis of the Obligations of all General-Purpose AI Model providers under the EU AI Act. We give a deep ~50 page, 300 footnote treatment and find so many tensions, loopholes, inconsistencies and more.

Link: files.michae.lv/papers/Veale...
The Obligations of Providers of General-Purpose AI Models Michael Veale1,2 and João Pedro Quintais 2 1Faculty of Laws, University College London 2Institute for Information Law, University of Amsterdam To appear in Zanfir Fortuna, Malgieri, González Fuster and Mantelero (eds.) The Artificial Intelligence Act — A Thematic Commentary (Hart forthcoming 2026). This version: November 2025. During the legislative process, the EU Artificial Intelligence (AI) Act was amended to include provisions related to general-purpose AI (GPAI) models. These broadly relate to transparency towards downstream users and relevant regulators, in addition to obligations connected to intellectual property. In this paper, we provide detailed analysis of these new provisions in the context of current technological applications and emerging trajectories, connecting them to computing literature and practice, and the broader context of connected and adjacent legal regimes, in particular copyright and relevant emerging case law. We find that there are a significant number of inclarities, tensions and contradictions both within the text, between the text and other legal regimes, and between the text and guideline documents, such as the Code of Practice on General-Purpose AI and recent guidelines by the European Commission. We identify a range of issues with the scoping of the provisions which may undermine its policy goals and create loopholes for regulatory avoidance, such as those relating to non-commercial models, open-source models, and model finetuning along the value chain. We find that the Code of Practice contains significant omissions and misstatements, some of which may present a compliance risk for an entity choosing to rely on the Code. We do not consider the provisions on GPAI models which present a systemic risk, which are dealt with elsewhere in the volume which this work will form a part of. A table of contents for the paper - 47 pages long, covering legislative background, definitions, information/documentation requirements, copyright, demonstration of compliance, exclusions and concluding remarks.
November 18, 2025 at 1:56 PM
michae.lv
uncanny timeline
Ellen De Generes oscar selfie, most people are changed but not hugely, their faces look a bit like an actor playing them
November 15, 2025 at 11:14 AM
michae.lv
cursed timeline
Ellen de generes Oscar selfie photo structure but people are dressed in normal looking clothing, faces very distored, somewhat like a horror movie
November 15, 2025 at 11:14 AM
michae.lv
The future that privacy advocates want:
A photo resembling the Ellen De Generes Oscar Selfie but everyone's faces are swapped with someone else often looking quite bad
November 15, 2025 at 11:14 AM
michae.lv
q: when you click on that link from the UK you get a country wide block. but you take UK subscribers, like me? why is the UK blocked?
November 14, 2025 at 4:31 AM
michae.lv
Apple Mail users: annoyed by Outlook spamy 'reactions'? I made a tool to prohibit ppl from reacting to your emails. Drag it into your applications folder; click then enable it in System Settings (it's not notarized); enable in Mail settings->Extensions. files.michae.lv/apps/badreaction.zip
October 30, 2025 at 11:31 AM
michae.lv
are you absolutely sure these are the small english channel boats you want to stop
map of d day landings
October 17, 2025 at 9:21 AM
michae.lv
would be great to have a call, I think we would all be very amenable! our undergraduate internet law final exam question this year screenshotted (after google spain excerpt)...
[41] It follows from all the foregoing considerations that (..) Article 2(b) and (d) of Directive 95/46 are to be interpreted as meaning that, first, the activity of a search engine consisting in finding information published or placed on the internet by third parties, indexing it automatically, storing it temporarily and, finally, making it available to internet users according to a particular order of preference must be classified as ‘processing of personal data’ within the meaning of Article 2(b) when that information contains personal data and, second, the operator of the search engine must be regarded as the ‘controller’ in respect of that processing, within the meaning of Article 2(d). Data protection sees a search engine as a controller, determining the means and purposes of the processing of personal data, while intermediary liability law sees a search engine as a passive host, without knowledge of illegal content and not liable for it. Are these regimes in conflict? If so, how, and if not, why not? END OF PAPER
September 23, 2025 at 2:51 PM
michae.lv
Hal Varian (later Google’s Chief Economist) in 2006. www.latimes.com/archives/la-...
“The first dose is free,” said Hal Varian, a professor of information management at UC Berkeley, facetiously comparing Microsoft’s anti-piracy policy to street-corner marketing of illicit drugs. “Once you start using a product, you keep using it.”
September 22, 2025 at 8:01 AM
michae.lv
We already had a fifth freedom, the Freedom of Movement of Non Personal Data. What, you hadn’t heard?
Free flow of non-personal data: Parliament approves EU’s fifth freedom Press Releases PLENARY SESSION IMCO 04-10-2018 - 12:16 New rules aimed at removing obstacles to the free movement of non-personal data within the EU for companies and public authorities were adopted by MEPs.
September 10, 2025 at 12:46 PM
michae.lv
first is from westminster (coming from west) and second is the way the bus is marked (towards hampstead heath). crash was in allington st, angled left. eyewitness must be wrong?
September 4, 2025 at 11:23 AM
michae.lv
you’ve missed out the section there on “standing authority” www.college.police.uk/app/armed-po... which is what applies here. up to chiefs, and while airport security plans aren’t public, airport police are part of the clear criteria. see doc below. this is a non-story.
September 3, 2025 at 7:11 AM
michae.lv
In a new digital twist to environmental individualisation of responsibility, the UK government is now advising people to “[d]elete old emails and pictures” in data centres to help with the current drought. www.gov.uk/government/n...
HOW TO SAVE WATER AT HOME Install a rain butt to collect rainwater to use in the garden. Fix a leaking toilet – leaky loos can waste 200-400 litres a day. Use water from the kitchen to water your plants. Avoid watering your lawn – brown grass will grow back healthy. Turn off the taps when brushing teeth or shaving. Take shorter showers. Delete old emails and pictures as data centres require vast amounts of water to cool their systems.
August 12, 2025 at 3:34 PM
michae.lv
This section is particularly weak.
August 12, 2025 at 3:27 PM
michae.lv
Has Oxford University Press validated whether such quizzes are pedagogically useful, or even accurate and faithful? Will they run them by the authors? Or is the suscription cost to Law Trove (hosts the most widely used UK legal UG textbooks, £50K+ a year per institution) not enough for that?
We’re excited to share our brand-new AI Study Assistant, developed in collaboration with Law lecturers and students to improve efficiency for students using Law Trove, enabling them to access quality, peer reviewed textbooks quicker than ever before: Input a query and the AI Study Assistant will rapidly navigate to content within Law Trove, connecting students quickly to our expert authors' authoritative research Students can ask the tool to formulate a quiz; it will then generate questions on a specific topic to consolidate their understanding The AI Study Assistant is included in all subscriptions to Law Trove; it will generate answers or quizzes based on the content to which a user has access. Everything entered into the tool is anonymous, and because only our trusted content is included, you can have confidence that your students are accessing reliable source material.
August 5, 2025 at 9:16 AM
michae.lv
training data
windows xp cloud and fields background
August 2, 2025 at 9:55 AM
michae.lv
i’m sure the trip was worth the 335 views none of which were just internal
video ‘a venentian delight with chef aldo zilli’ with picture of sardines, 335 views, subtitle on video ‘thank you to the university of warwick for bringing me here’.
July 31, 2025 at 8:33 AM
michae.lv
Warwick hired “celebrity chef Aldo Zilli to create a “sardine and prawn” dish to celebrate the 700th anniversary of Venice”, plus more Ajay Teli trips, talks - as if future students care about a CMO! This isn’t appealing to students—it’s maladministration www.timeshighereducation.com/news/warwick...
July 31, 2025 at 8:24 AM
michae.lv
New AU 🇦🇺private members bill (Criminal Code Amendment (Using
Technology to Generate Child Abuse
Material) Bill 2025) criminalises child abuse generation tech (any tech w that as ‘sole/dominant purpose) creation/distribution similarly to UK Crime & Policing Bill parlinfo.aph.gov.au/parlInfo/dow...
July 30, 2025 at 3:35 PM
michae.lv
OpenAI insists cheating at universities is something ‘holistic’ for (people other than them)
solve while refusing to implement text watermarking because a survey showed people would use competitors if they did. www.timeshighereducation.com/news/chatgpt...
July 30, 2025 at 7:25 AM
michae.lv
hand clasp meme showing 'paleo' and 'ultraprocessed food' clasping hands over the topic of 'neanderthal saturated fat machine'
July 29, 2025 at 3:14 PM
michae.lv
for reading on this see our UCL module on Sexuality and the Law’s week on age verification www.homepages.ucl.ac.uk/~ucqnmve/syl...
see link for screen reader version! see link for screen reader version!
July 26, 2025 at 5:42 PM
michae.lv
Anthropic do conversation clustering using their 'Clio' system but misleadingly call it 'privacy preserving' and 'anonymised'. If you read the paper it is clear that clusters remain linked to individual accounts for investigation making them clearly *not* anonymised. www.anthropic.com/research/clio
real cases of such behavior identified by Clio. In all of these cases, when a suspicious cluster was discovered, it was reviewed by a small number of designated members of our Trust and Safety team who are authorized to manually review conversations—in a secure environment, and under strict privacy controls—that have been flagged for violating our Usage Policy.6 1. Using automated accounts for search engine optimization. Clio identified a large cluster of conversations in which Claude was asked to generate keywords for search engine optimization about the same topic across many different accounts. While none of the individual conversations violated our Usage Policy, after investigating the accounts further, we determined that they were engaged in coordinated abuse (which is against our Usage Policy) and we removed them from our systems. 2. Using automated accounts for explicit content generation. Clio identified a large cluster of conversations from many different accounts that used an identical complex prompt structure to engage Claude in sexually explicit role-play. After investigating the accounts further, we determined that they were coordinated and systematically violating our Usage Policy, and we removed them from our systems.7 3. Reselling access to Claude in violation of our terms and policies. Clio revealed that a large amount of traffic from several accounts suggested violations of certain policies. On further investigation, we determined that these accounts were reselling unauthorized access to Claude in violation of our Usage Policy, and so we removed the violative accounts from our systems. How Clio works: Privacy-preserving analysis at scale Traditional, top-down safety approaches (such as evaluations and red teaming) rely on knowing what to look for in advance. Clio takes a different approach, enabling bottom-up discovery of patterns by distilling conversations into abstracted, understandable topic clusters. It does so while preserving user privacy: data are automatically anonymized and aggregated, and only the higher-level clusters are visible to human analysts.
July 26, 2025 at 4:05 PM