Superintelligence Europe — No. 012

The UK’s Department for Science, Innovation and Technology is circulating proposals designed to persuade Anthropic — the San Francisco-based AI company behind the Claude chatbot — to significantly expand its London presence and potentially pursue a dual stock listing on the London Stock Exchange. The plans will be presented to CEO Dario Amodei during his visit to the UK in late May, when he is scheduled to meet European customers and policymakers. Prime Minister Keir Starmer’s office has backed the effort. London Mayor Sadiq Khan has separately written to Amodei urging the company to commit more deeply to the city, calling London a “stable, proportionate and pro-innovation environment.”

The UK move is a direct play on Anthropic’s highly public dispute with the US Defense Department. The Pentagon blacklisted Anthropic as a national-security supply-chain risk after the company refused to allow its Claude AI to be used for US military surveillance operations or autonomous weapons. A San Francisco federal judge temporarily blocked the blacklisting, and Anthropic has a second lawsuit pending. The company — backed by Amazon and Alphabet, valued at around $380 billion — already employs approximately 200 people in the UK, including researchers, and counts former Prime Minister Rishi Sunak as a senior adviser. It is in preliminary discussions with Goldman Sachs, JPMorgan Chase, and Morgan Stanley about underwriting an IPO as early as October 2026. One person familiar with the DSIT proposals described a dual UK-US listing as “the dream,” while acknowledging it as unlikely. Business Secretary Peter Kyle confirmed to the FT that Anthropic is among the fast-growing companies he wants to invest more in the UK.

UK Research and Innovation published the results of its independent midterm review of the Alan Turing Institute — the UK’s national institute for data science and artificial intelligence — and the verdict is sharp: overall strategic alignment and value for money are “not yet satisfactory.” The review, while acknowledging the institute’s scientific excellence and strong research foundations, found that it has failed to articulate a clear strategic purpose and has spread its resources too thinly across fragmented academic interests overlapping with other public institutions. UKRI is now demanding significant structural change and has set a September 2026 deadline for the institute to submit a transformation plan, which will then be independently assessed.

The recommended new direction is unambiguous: a “clear, single purpose focused on national resilience, security and defence.” This is the culmination of pressure that has been building since July 2025, when Technology Secretary Peter Kyle demanded an overhaul and called for the institute to focus on defence and security rather than its broader academic mandate. The chair, Doug Gurr, resigned on 1 April 2026 to become chair of the Competition and Markets Authority. His interim replacement is Vanessa Lawrence. New CEO George Williamson — previously head of His Majesty’s Government Communications Centre — starts in May 2026, deepening the defence pivot. UKRI’s funding of £100 million to the institute is explicitly conditional on the September plan being delivered.

With Hungary’s parliamentary election on April 12, the country is now one week from what Carnegie Endowment researchers describe as the most consequential vote in fifteen years of Orbán rule — and Europe’s first election in which AI-generated synthetic media has been deployed at industrial scale. Prime Minister Viktor Orbán’s ruling Fidesz party published an AI-generated video in February depicting a Hungarian soldier in uniform, kneeling blindfolded on a battlefield, being executed — captioned to suggest this is the future if opposition leader Péter Magyar wins and is “dragged into the Ukraine war.” Orbán’s chief of staff, Gergely Gulyás, did not deny the video was AI-generated when asked at a press briefing. It was confirmed by Reuters to have been made using Google’s AI models.

That video is not isolated. The pro-Orbán National Resistance Movement has spent over €1.5 million on unlabelled AI-generated videos targeting Magyar on TikTok, Facebook, and Instagram. A separate AI-generated video depicted a fake phone call between European Commission President Ursula von der Leyen and Magyar, ostensibly discussing financial aid to Ukraine. EU DisinfoLab has documented a coordinated foreign information manipulation campaign on TikTok linked to Russia’s “Matryoshka” operation, using synthetic news anchors and deepfake celebrity endorsements to amplify pro-Orbán narratives. Despite the campaign, Magyar’s Tisza party leads Fidesz by 8–12 points in most polls. His posts on social media receive twice the engagement of Orbán’s.

Vision Compliance — a Zagreb-based European regulatory advisory firm — published its 2026 EU AI Act Readiness Analysis on April 1, with full coverage circulating across European policy and compliance communities on Sunday. The headline finding from assessments conducted across financial services, healthcare, technology, manufacturing, energy, retail, telecommunications, and transport: 78 percent of organisations have not taken meaningful steps toward AI Act compliance. The most common failure mode is awareness without action — companies know the regulation exists, but very few understand what it actually requires at the operational level.

The compliance gaps Vision Compliance documents are structural, not superficial. Most organisations lack AI system inventories entirely: without knowing which AI systems they operate, risk classification is impossible. The AI Act introduces requirements that go significantly beyond data protection — including conformity assessment procedures, post-market monitoring obligations, and technical documentation standards — that are entirely new territory for even GDPR-mature compliance teams. The report notes that GDPR-compliant organisations are better positioned on data governance, but that advantage does not extend to the Act’s new requirements. One finding stands out: the August 2, 2026 enforcement deadline for high-risk AI systems is approaching whether the Digital Omnibus’s proposed extensions are agreed by trilogue or not. Organisations that assume the extension will materialise are taking regulatory risk.

TNW’s Editor-in-Chief framed the week as evidence of a single structural instinct running through European capital: “build the infrastructure layer first.” The week ran from Mistral’s $830M debt raise at one end to a €1.1M workpod pre-seed at the other — a useful reminder of how wide the band of European ambition now runs. The confirmed highlights:

Credible accounts and publications driving Sunday’s European AI conversation. Filtered for genuine signal.