iiinova — where your digital self keeps evolving

michael veale @michae.lv · 21/08/2026

disappointed that out of the huge number of digital sovereignty conferences, none have yet centred the lived experience of actual sovereigns

3121

michael veale @michae.lv · 18/08/2026

thanks! I will fix this, it's a bit of a mess because Curia doesn't have an API, and joined cases mess things up a bit, but you can actually reverse engineer their interface...

it works but all text that has less entropy like code requires a longer passage for reliable detection

030

Sainsbury’s store pauses AI scanning after false shoplifting accusation

sainsbury‘s pauses AI shoplifting detection face scanning from Gordon’s Wine Bar spin out surveillance company Facewatch after incorrect customer ejection www.theguardian.com/technology/2...

theguardian.com

Supermarket chain says ‘human error’, not its Facewatch technology, to blame for ejecting a customer

1134

when google says it 'open sourced' synthid, it did (and had a paper in nature some time back) but it of course didn't open source its specific private key

021

have not seen anyone note that, in relation to AI SynthID text watermarking: all methods deployed by AI firms rely on private keys for security so no cheap/local/private detection. EU AI Act guidelines say it should not be this way, it should be local/interoperable. the deployed methods are not!

2133

Emanuel Maiberg @emanuelmaiberg.bsky.social · 17/08/2026

given Amazon provide the 'Sidewalk' infrastructure that makes Tile trackers work via Alexa and Ring devices as detectors, they don't exactly have a leg to stand on when investigative journalists slip an AirTag in a book

061

Reposted by michael veale

We put a tracking device in a shipment of rare books acquired by an anonymous buyer. It ended up at an Amazon facility where the company scans books for training data and destroys them in the process: www.404media.co/we-tracked-a...

404media.co

We Tracked a Shipment of Rare Books. It Ended at an Amazon AI Training Facility

We placed a tracking device in a shipment of rare books to see which AI company was buying it, and found an Amazon facility where Amazon scans and destroys books.

391572914

the EU AI Act's requirement in the guidelines that systems are interoperable, open, run locally etc, is the biggest issue with current designs, and one I haven't seen anyone pick up on

the detector code may be open sourced but SynthID relies on a private key

google has also been doing text watermarking for gemini for months and months, just never released a detector

141

yes, this blog is v wrong, a shame because he's said some sensible & useful things in the past especially regarding apple app store policies & politics. ai watermarking interestingly invented by visiting researcher at OpenAI many years ago; co built it then sat on it as scared it would hit adoption

190

Details on the data sources and citing countries covered here homepages.ucl.ac.uk/~ucqnmve/law... with the most data from France, Germany, UK, EU, Ireland, Estonia, Austria, Netherlands, Australia, Belgium, Finland, Slovakia, ECHR, NZ & Sweden, and admin decisions/guidance from others.

homepages.ucl.ac.uk

Sources

michael veale @michae.lv · 14/08/2026

I've updated UCL Massively Crossreferenced Digital Laws — it now features ~40 EU instruments as well as Irish & UK digital laws; what cites each article/para/sub para from a db of 6.5M cases, statutes, decisions, guidance, pending CJEU preliminary rulings & more homepages.ucl.ac.uk/~ucqnmve/law...

Image of the linked interface focussing on Art 59 of the AI Act

11410

@pieterhaeck.bsky.social why not mention the california law that actually applies to on-market models for watermarking that came into force on 2 aug with very similar provisions (although not for text, only for other media?). brussels isn't the only regulator & omnibus has delayed this provision too

Have added and parsed 500,000 German Länder judgments today to the citing pool today. your köln one of them, and EECC now features annotations by it. if you see more missing things that should be there, let me know!

*california aligned its with europe

AI act deadline WAS august 2, which is why california aligned it’s with anthropic, so anthropic was likely working towards this deadline anyway with EU in mind. But as of today they do not have to face liability in Europe for not watermarking, not until later this year.

Code of practice signature is a voluntary instrument.

i see government is taking its online presence seriously in this troubled digital age, passing GDS responsibility to the minister in charge of ‘ceremonials’ (ie remembrance day parades) and tourism.

011

This isn’t correct. The AI Omnibus delayed implementation of this for existing providers whose products were already on the market. This is Anthropic complying with Californian AI watermarking law (SB 942), which came into effect 2 Aug.

120

so *technically* the footnote should say 'doctoral', if we give such benefit of doubt

000

if these numbers are true, I suspect both relate to part time EdD programmes (cambridge, ucl, bristol, glasgow, OU...) rather than PhD ones, many of which are only 40-50K words total and supervisions can rack up due to part timers going on for ages

100

michael veale @michae.lv · 10/08/2026

not yet, the underlying database only has federal courts in from official government apis (draws on the RII and Neuris apis). no official apis exist for german länder courts to my knowledge. it’s on my list to integrate from de.openlegaldata.io which would then include more german courts

100

the whole 'Acts' thing makes for a load of idiosyncracies when trying to match irish pincites to national data protection law...

Jonty Wareing @jonty.bsky.social · 10/08/2026

Reposted by michael veale

I always suspected a UK government would eventually reverse course on "coding in public". So I've been backing up all government code. For six years. Today I finally generated a report on what has been deleted. Its much worse than I realised. github.com/uk-gov-mirro...

Deleted UK Government code repositories

Total deleted repositories: 1526
Organisations affected: 56

See RECENTLY_DELETED.md for the most recently deleted repositories.

Organisation Deleted Repos % Deleted
hmrc 195 9.8%
nhsdigital 169 33.8%
SkillsFundingAgency 154 22.7%
informatics-lab 154 100.0%
UKHomeOffice 101 7.5%
ministryofjustice 74 2.8%
ONSdigital 66 4.3%
moj-analytical-services 65 75.6%
ukwa 47 46.1%
dwp 46 6.7%
UKGovernmentBEIS 43 27.4%

6219134

michael veale @michae.lv · 10/08/2026

how much do you think is them being scared (rationally or not) of someone searching for vulnerabilities?

140

michael veale @michae.lv · 10/08/2026

have added some irish instruments now (at the bottom of the page), still working on increasing coverage but would be useful sensecheck if you see any you think should be cited more by court cases than they are homepages.ucl.ac.uk/~ucqnmve/law...

homepages.ucl.ac.uk

UCL Massively Crossreferenced Digital Laws

230

Inside the Black Box @itbb.bsky.social · 07/08/2026

it's less about the liability, it's more that cloudflare's actual policy is to block scrapers it detects are not adhering to robots.txt files that it parses.

120

Reposted by michael veale

The default is the consequential part. Scraping consent has operated as opt-out-if-you-knew-to, so silence became permission. Cloudflare now turns silence into reservation at a chokepoint covering over 20% of domains, changing the baseline for sites that never expressed a view.

021

it's a good question, but insofar as cloudflare itself is enforcing it, it's a hell of a lot more enforcable than robots.txt alone ever was

meta fears a google 'glasshole' style catchy moniker situation for their unpleasant new tech. perhaps it's time to strike fear into their corporate partner and make 'raybanned' a verb?

47425

they are also attempting to extend legal interpretation (potentially) with their 'use=reference' category, which their docs indicate is says allows you to index, make an excerpt and link back, but does NOT give permission to summarise.

1102

this is quite a big deal and ramps up potential liability. it also is a move to put pressure on Google to split Googlebot into search and training (allowing people to block training but allow search).

1193

michael veale @michae.lv · 06/08/2026

Cloudflare (>20% of all domains) has started automatically implementing DSM Dir art 4 rights reservation (ie copyright/TDM objection to scraping) for all sites that do not have a robots.txt file. They also adding purpose fields to robots.txt unilaterally. developers.cloudflare.com/bots/additio...

# As a condition of accessing this website, you agree to abide by the following
# content signals:

# (a) If a Content-Signal = yes, you may collect content for the corresponding
# use.
# (b) If a Content-Signal = no, you may not collect content for the
# corresponding use.
# (c) If the website operator does not include a Content-Signal for a
# corresponding use, the website operator neither grants nor restricts
# permission via Content-Signal with respect to the corresponding use.

# The content signals and their meanings are:

# search: building a search index and providing search results (e.g., returning
# hyperlinks and short excerpts from your website's contents). Search does not
# include providing AI-generated search summaries.
# ai-input: inputting content into one or more AI models (e.g., retrieval
# augmented generation, grounding, or other real-time taking of content for
# generative AI search answers).
# ai-train: training or fine-tuning AI models.
# use: how AI systems may consume the content (immediate, reference, or full).

# ANY RESTRICTIONS EXPRESSED VIA CONTENT SIGNALS ARE EXPRESS RESERVATIONS OF
# RIGHTS UNDER ARTICLE 4 OF THE EUROPEAN UNION DIRECTIVE 2019/790 ON COPYRIGHT
# AND RELATED RIGHTS IN THE DIGITAL SINGLE MARKET.

# BEGIN Cloudflare Managed content

49143

@tjmcintyre.com honorary irish citizen please, i’m one generation out (perhaps can be joined with maybe one day if collected causing a €400m windfall for the Irish treasury, though not sure if the gov wants it…) ww2.courts.ie/view/Judgmen...

michael veale @michae.lv · 06/08/2026

@jenitennison.com i, too, turned off my firewall and exposed all my ports to the internet. in doing so i have shown the dark reality of how vulnerable we all are. something must be done. please buy my network services.

michael veale @michae.lv · 05/08/2026

also AI security institute: "please don't make us look for anything that would endanger our relationship with these big companies who give us their apis on a purely voluntary basis!"