Superintelligence Europe — No. 017

The UK government tabled an amendment to the Crime and Policing Bill on Friday that would make senior technology executives personally criminally liable if their platforms fail to comply with Ofcom’s enforcement decisions to remove non-consensual intimate images. Under the proposed law, executives without a reasonable excuse could face imprisonment, a fine, or both. The amendment will be debated in the House of Commons next week and represents the most direct escalation of individual executive accountability in British platform regulation since the Online Safety Act came into force.

Technology Secretary Liz Kendall framed the move explicitly in terms of AI-generated abuse. The amendment builds directly on the government’s January 2026 legislation criminalising the creation of non-consensual intimate images — a response to the surge of AI-generated deepfake sexual content on X’s Grok tool and across other platforms. In February, the government required platforms to remove reported non-consensual intimate images within 48 hours. Friday’s announcement goes further: it places the compliance obligation on named executives, not just the platform, with criminal consequences for failure.

Germany’s Handelsblatt reported Friday, citing sources, that OpenAI and its ChatGPT chatbot are set to be classified as a Very Large Online Search Engine (VLOSE) under the European Union’s Digital Services Act. Reuters carried the wire. OpenAI declined to comment when contacted by Handelsblatt. A spokesperson for the European Commission told Handelsblatt that the available user data was being reviewed. OpenAI itself had reported that ChatGPT’s search feature reached an average of 120.4 million monthly EU users over the previous six months — a figure 2.7 times above the 45 million threshold that triggers VLOSE designation obligations.

VLOSE designation under the DSA would bring ChatGPT under the EU’s most demanding regulatory tier, a set of obligations that go significantly beyond what the AI Act alone requires. OpenAI would be required to conduct annual systemic risk assessments covering impacts on civic discourse, electoral processes, mental health, and the protection of minors. It would be required to provide transparency on how model outputs are moderated, give researchers access to data, and implement risk mitigation measures whenever it deploys a new functionality likely to have a critical impact on systemic risks. Unlike the AI Act, which focuses on the model itself, the DSA focuses on how it operates as a platform intermediary — the systemic role ChatGPT now plays in how hundreds of millions of people access and process information.

The European Society of Cardiology published a press release on Friday confirming progress on the EU’s NextGen project — a Horizon Europe-funded initiative working to remove the data integration barriers that have prevented AI cardiovascular models from reaching their clinical potential. The core problem NextGen addresses is structural: health data across Europe exists in incompatible formats, governed by different national privacy frameworks, stored in systems that cannot communicate with each other. Genomic data is particularly complex — information-rich, individually identifiable, and held in formats that do not integrate with clinical records or imaging data. NextGen is building the “digital fabric” that allows all three data types to be combined securely and used to train the next generation of cardiovascular AI models.

The project involves a 21-member consortium including the European Society of Cardiology, Queen Mary University of London, the Earlham Institute, UMC Utrecht, and institutions from Germany, France, Finland, Italy, Switzerland, and the US. Its tools ensure health data remains meaningful and readable across different borders and hospital systems without losing clinical context, and allow researchers to discover relevant cardiovascular datasets without moving or exposing patient information. Governance is hard-coded into the data architecture itself. Five clinical pilot sites are running real-world demonstrations. The project is funded by €7.6 million from Horizon Europe plus additional grants from the Swiss State Secretariat for Education and UK Research and Innovation.

Bloomberg’s framing was the most important addition Friday brought: OpenAI is pausing Stargate UK as it reins in spending ahead of a public listing anticipated as early as Q4 2026. The company closed a $122 billion funding round at an $852 billion valuation in late March. Companies approaching an IPO tighten capital allocation, avoid open-ended international commitments, and reduce exposure to projects with uncertain timelines. The pause fits that pattern. The UK is not uniquely problematic — it is one of several locations where conditions do not justify a pre-IPO capital commitment.

TNW’s analysis surfaced the most concrete structural obstacle: UK grid connection requests surged from 41 gigawatts in November 2024 to 125 gigawatts by June 2025, with approximately 75 gigawatts attributable to data centre projects. Grid connections take three to eight years; data centres take 18 to 24 months. The Cobalt Park AI Growth Zone designation provides streamlined planning — but not a shorter grid queue. UK industrial electricity costs more than four times equivalent US locations. Both barriers are structural, not policy resolvable in the short term.

Hungary votes on Sunday 12 April. Viktor Orbán’s Fidesz faces Péter Magyar’s TISZA party, with polls consistently showing Magyar at 58% to Fidesz’s 35%. The campaign has deployed AI-generated political content at a scale not previously documented in an EU member state election. NewsGuard confirmed a 34-account TikTok bot network — 22 accounts created within two days in January 2026 — generating approximately 10 million views of AI-generated videos boosting Orbán and discrediting Magyar. TikTok confirmed the accounts constitute a covert influence operation. EDMO fact-checkers documented AI-generated news anchor reports about Magyar with 100% AI-detection certainty. A deepfake execution video and a fabricated von der Leyen phone call were distributed by Fidesz. The Russian Matryoshka disinformation network is confirmed active on X and Telegram.

Friday brought five stories that appear separate but read together as one: Europe is accelerating the enforcement of AI accountability across multiple dimensions simultaneously. The UK is threatening prison for platform executives who allow AI-generated intimate abuse to persist. The EU is reviewing whether to bring ChatGPT — now at 120 million EU monthly users — under its toughest regulatory tier. Hungary is holding an election in which AI-generated disinformation is documented and confirmed, and which will test whether the EU AI Act’s grey zone is politically tolerable after the results are known. The ESC published a Horizon Europe project on Friday showing what genuine AI-for-public-good looks like: federated, privacy-by-design, clinically grounded, and built by a 21-country consortium.

The week that started with the EU AI Continent’s one-year milestones ended with a UK company criminalising executive inaction on AI-generated abuse, an EU regulator reviewing whether to classify the world’s most used AI chatbot as a systemic platform risk, and an election in which AI disinformation infrastructure operated without domestic enforcement consequence. This is what the deployment phase of the AI governance cycle looks like.