Mythos breached, Merz's ultimatum, OpenAI EMEA, Nicosia

Anthropic is investigating a report that a private Discord group gained unauthorised access to Claude Mythos Preview through one of the company’s third-party vendor environments. According to Bloomberg, the group gained access on the same day Anthropic announced the model’s limited launch weeks ago, and has been using it regularly since — though not for malicious purposes.

Mythos, described by Anthropic as vastly outperforming other AI models in vulnerability detection and exploitation, was only authorised for a closed group of companies including Apple, Google, Microsoft, Nvidia, and JP Morgan Chase. UK financial institutions are set to begin using Mythos this week. Japan’s Finance Minister is meeting banks today to discuss the model’s risks. Canada is also in active discussions.

Ireland’s National Cyber Security Centre director Richard Browne told an Oireachtas Joint Committee last week that the technology would be “in the hands of bad actors within months.” The leak suggests that timeline may already be compressing.

At Hannover Messe, Siemens CEO Roland Busch told Bloomberg that most of the company’s €1 billion industrial AI investment will be directed to the United States, not Europe, because the EU’s AI Act and Data Act treat industrial machine data the same as personal consumer data. For a company already operating under sector-specific regulations, this creates redundant oversight that cannot be justified to shareholders.

German Chancellor Friedrich Merz used the same stage to call for a formal exemption for industrial AI from what he called a regulatory framework “too tight for AI within the European Union.” This is a rare, direct public intervention by a German Chancellor against a specific Commission regulatory regime — timed precisely five days before the Omnibus trilogue.

Siemens is Germany’s most valuable listed company at €194 billion market capitalisation. The political and commercial lobbying is now unified and public.

OpenAI has appointed Emmanuel Marill as its first-ever EMEA Managing Director, announced today. Marill previously led Airbnb’s EMEA, Australia, and New Zealand operations, and before that served as Meta’s financial services and automotive industry lead for Facebook and Instagram. He will be based in Paris and report directly to Chief Strategy Officer Jason Kwon.

Marill will collaborate with OpenAI’s EU headquarters in Dublin, which currently employs approximately 80 people. The appointment reflects what OpenAI called “strong momentum in EMEA,” with weekly active ChatGPT users in the region growing 70% year-on-year. Germany, France, UK, and Spain are among OpenAI’s top EMEA markets for ChatGPT and Codex.

The timing is significant: OpenAI is navigating increasing EU regulatory pressure and business resistance over digital sovereignty — challenges that a dedicated EMEA leadership structure is now expected to address directly ahead of a likely IPO.

All 27 EU heads of state are arriving today in Cyprus for a two-day Informal European Council hosted by President Nikos Christodoulides under the Cypriot Council Presidency. Roads around the Nicosia conference centre have been closed since 07:30 CET. The working sessions open this evening at Ayia Napa Marina.

The formal agenda covers geopolitical environment and the EU multiannual financial framework 2028-2034. But Cyprus President Christodoulides confirmed to AP today that leaders will also discuss Article 42.7 of the EU Treaty — the mutual defence clause — and what happens if a member state calls for EU-wide military assistance. A working lunch with Middle East regional partners follows on April 24.

The timing cannot be separated from the regulatory calendar. Merz, who just called for an industrial AI carve-out at Hannover Messe, is attending this summit. The political conversations that begin today in Nicosia will shape the negotiating positions that arrive in Strasbourg in five days.