There's a lot of pressure to drastically increase Europe's compute capacity, from CADA to the Europe 2031 scenario. The assumption underneath: building more capacity buys more independence. Our new piece with @techpolicypress.bsky.social takes that apart.

On the heels of European Anthropic-Fable scare, the illustrious @frederike.bsky.social and I have a new piece out on the potential dangers on tethering the European AI sovereignty hopes on brilliance at the application layer. Happy to hear comments!

Back in 2022, France and Germany sought to extend the mandate of the European Public Prosecutor’s Office to cover violations of EU sanctions, but their efforts failed to gain sufficient support. A new opportunity has now emerged, writes Elie Cavigneaux. www.swp-berlin.org/en/publicati...

From using Europeans' pensions to build frontier models, to allowing hyperscalers to build all the data centers in exchange for frontier model access - I've heard mad takes these past few days 😱 Read our AI Sovereignty in a Captured Ecosystem series ! www.techpolicy.press/europes-ai-s...

Great piece by @frederike.bsky.social & @leevisaari.eurosky.social on the risks of building "European AI" on top of foreign models. One thing I'd add: digital decade goals indirectly incentivise member states to ignore it in pursuit of AI adoption and "unicorns". www.techpolicy.press/europes-ai-s...

Germany's DeepL builds on AWS. France's Mistral builds with Nvidia. Sweden's Lovable just got a jump scare as Anthropic moved in to eat its lunch. That's the starting point of a new series we're launching with @leevisaari.eurosky.social @techpolicypress.bsky.social @ainowinstitute.bsky.social

I have a lot of thoughts on the leaked EU Sovereignty Package...

I wrote about Iran's protests for @theatlantic.com: how AI manipulation and the suspicion of it is being weaponised to dismiss real footage of protests. This piece draws on work from my stellar team at WITNESS and years of studying information controls in Iran and beyond.

in the wake of us v. google, i wrote for @techpolicypress.bsky.social how generative AI is not going to magically unseat a dominant firm like google: it's going to enable it to become even more dominant. news in the past 24 hours shows exactly this (see thread!) www.techpolicy.press/decision-in-...

This needs more widespread coverage: The US has sanctioned five individuals, including the founder and Director of @hateaid.org - a German NGO that supports victims of online hate and harassment and image-based sexual abuse

"US companies have sold $1.7tn of investment-grade bonds in 2025, a near-record sum stoked by a rush of borrowing to fund AI infrastructure that has spurred concerns over a debt glut."

Re-up on this in light of the @404media.co story out today.

In response to WIRED’s request for comment about users being able to generate bikini deepfakes with ChatGPT, a spokesperson for OpenAI claims the company loosened some ChatGPT guardrails this year around adult bodies in nonsexual situations.”

George Osborne getting a new job isn't exactly news. But him being the latest former politician to join a tech company pushing out AI into government absolutely is. My latest for @theguardian.com www.theguardian.com/commentisfre...

💀

Here's how this will go down: 1) Adversarial attacks on Sora will be unrelenting, putting Disney characters in all kinds of compromising videos 2) These will be a new firehose of slop that fly-by-night content farms will monetize on Facebook & elsewhere 3) Fan backlash will be significant

Verheerender und vielsagender Satz von Kanzler Merz zur Unterwanderung des europäischen Zusammenhalts durch die US Regierung unter Trump, gerichtet an die USA: „Wenn ihr mit Europa nix anfangen könnt, dann macht wenigstens Deutschland zu eurem Partner." Es ist nicht zu fassen.

Teach the environmental harms of AI and its infrastructure.

Macron this week proposed trading away export controls on European made semiconductor equipment. Trump just weakened the US arguments against that.

Black women told you this in 2018.

This was just sent to me by an LAUSD parent. A 4th grader was assigned to design a book cover for Pippi Longstocking using Adobe for Education. Here is what the AI tool generated:

A whopping 65% of Germans say Germany is acting too cautiously in asserting its interests towards the United States.

120m is no deterrent to X. Musk will moan in public - in private, he will be doing cartwheels. Tech compliance theatre won't cut it anymore. Europeans want enforcement with teeth - up to revoking access to the Single Market for the worst actors. And Musk + X are the slop merchants par excellence.

Bei der «Stadtbild»-Debatte ging es stets um Bilder, um rassistische Phantasmen der Reinigung wie um die nostalgische Sehnsucht nach einer Vergangenheit, die nur in Bildern existiert. AI Slop nimmt nun im Imaginären vorweg, was man für die Wirklichkeit plant: die ethnische Säuberung der Innenstädte

A few thoughts on the new US national security strategy that was released today. 🧵 www.whitehouse.gov/wp-content/u...

🚨 A huge leak exposes the new targets and internal operations of Intellexa, the secretive and murky company behind the notorious Predator spyware. Introducing #IntellexaLeaks, a joint investigation with partners @insidestory.gr, @haaretzcom.bsky.social & WAV Research Collective 🧵👇

“This incident shows how Google is trying to leverage its Search monopoly into a monopoly on answers themselves” www.bleepingcomputer.com/news/artific...

Fascinating:

'Small changes to the tone of posts fed to users of X can increase feelings of political polarisation as much in a week as would have historically taken at least three years, research has found.' 1/3

Theft on top of theft.

When the Trump administration demanded ICEBlock’s removal, Apple complied—no rival app stores, just one gatekeeper’s decision, explains Meher Sethi. The takedown exposed how easily government pressure can travel through Apple and Google’s app-store duopoly and the risk it poses to democracy.

Multiple surveys are telling the same story: AI adoption has stalled. Census data shows ~11% of workers use AI, with rates falling at 250+ employee firms. Surveys show usage spiked then plateaued. Meanwhile Big Tech is planning $5T in AI spend by 2030 that needs $650B/year in revenue vs ~$50B today

So thrilled to finally have @ketanjoshi.co on the podcast! We dig into what the numbers are telling us about the climate impact of data centers — and it’s not a good story. These tech companies need to be held accountable for the climate cost of their AI ambitions!

7️⃣ years ago, 7 of our members filed complaints against Google over its tracking of users’ locations. A lot can happen in 7 years: the GDPR was reformed, Europeans born in 2018 are now in primary school, and the people below joined our digital & enforcement teams! Yet, this case is still pending⏱️

Im Europäischen Parlament stimmen die Konservativen mit rechts außen. Die Stabilität, auf der die Union basiert, wird so aufgegeben. Die politische Mitte zerlegt sich.

This story should be covered waaaaayy more than it is. The US has sanctioned six judges if the international criminal court ICC, because Trump didn’t like them issuing arrest warrants for Benjamin Netanyahu and former Minister of Defense Yoav Gallant. www.lemonde.fr/en/internati...

when its definitely not a hostage situation

Germany has the most data centres in Europe — but the boom is pushing its energy grid to breaking point. In hotspots like Frankfurt, power is already maxed out. Read our new investigation with @algorithmwatch.org. www.techpolicy.press/germanys-dat...

Although the scale of this is still to be assessed, this is the inevitable outcome of the monetization of the algorithm. Rage-bait becomes a personal revenue model. Nothing about this should be a surprise.

There is a whole underworld of exploited workers who make up the foundation of digital tech. We talk about content moderators, but content farms are also very exploitative industries where moderately educated people with reasonable English skills in the global majority are used to push content.

This.

this isn't 1 problem caused by 1 platform/piece of technology. it's many overlapping problems caused by a bunch of systems that were optimized for scale and/or built & rolled out without any real care for how they'd be used and abused. one can only conclude this is the desired use for all this tech

The company essentially turned a dial that made ChatGPT more appealing and made people use it more, but sent some of them into delusional spirals. OpenAI has since made the chatbot safer, but that comes with a tradeoff: less usage.

I want the wholesome fish tank Internet back 😭

The Netherlands talks a big game about tech sovereignty, but is now about to allow the domestic cloud provider that underpins its national digital identity system and several other critical government services to be sold off to a new American owner. Incomprehensible. nos.nl/artikel/2591...

Goddamn dude, just go ahead knock 25 percent off my 401k and get on with it, I can't take much more of this.

A US federal judge ruled on Tuesday that Facebook-owner Meta is not a monopolist in a case brought by the Federal Trade Commission over its acquisitions of Instagram and WhatsApp. Tech Policy Press associate editor @viacristiano.bsky.social looks at some of the key arguments in the ruling.

"What we see with media organisations is that they have to go to YouTube and TikTok to find their audiences but then they get shadowbanned when they mention genocide." — @sandervdwaal.bsky.social #EuroskyLive

Picture someone being sucked into a delusion spiral by chatgpt, and then ordering suicide equipment with it *mid-conversation* while it cheers them on "I just bought you $400 worth of garden rope, king"

Essential reading in the AI bubble and the wild Wild West that is the neocloud market - with lots of insight from @leevisaari.bsky.social www.theverge.com/ai-artificia...