The Signal — April 10, 2026
The Treasury Secretary and the Fed Chair don't usually call emergency meetings about an AI model, but that's what happened this week.
Bessent and Powell Summon Bank CEOs Over Mythos Cyber Risk
Treasury Secretary Scott Bessent and Fed Chair Jerome Powell convened an urgent meeting with Wall Street's top executives on Tuesday to discuss cybersecurity risks posed by Anthropic's Mythos model. The model can reportedly identify and exploit vulnerabilities across every major operating system and web browser.
CEOs from Citigroup, Morgan Stanley, Bank of America, Wells Fargo, and Goldman Sachs all showed up. JPMorgan's Jamie Dimon was unable to attend. Mythos access is currently limited to around 40 tech companies, including Microsoft and Google, and Anthropic proactively briefed government officials before release.
This is the first time senior financial regulators have convened banks specifically over what an AI model can do — not over market volatility or a hack that already happened, but over a model's potential, and that distinction is worth sitting with. The financial system's leadership decided this couldn't wait for a memo.
OpenAI Shelves Stargate UK
OpenAI has paused its Stargate UK data center project, citing high energy costs and regulatory headaches. The project was part of a much-hyped UK-US AI deal from September 2025, where American companies committed a collective £31 billion to British AI infrastructure. The word "committed" is doing heavy lifting here.
The delivery partner, Nscale, has never built a data center before and was supposed to deliver a supercomputer by 2027. Right now, the site is a scaffolding yard in Essex. A previous Guardian investigation found that many of the UK AI investment deals were "phantom investments," and this pullout is drawing cross-party criticism.
Context check: OpenAI just closed a $122 billion funding round at an $852 billion valuation in late March. They're not short on cash. The UK project isn't dying because OpenAI can't afford it; the economics just don't work, and maybe never did. When the biggest AI company on earth walks away from a flagship international deal, it tells you something about the gap between AI investment announcements and AI investment reality.
Sources: The Guardian, CNBC, The Register
Anthropic Sent Mythos to a Psychiatrist
Anthropic's 244-page system card for Mythos reveals something unprecedented: they hired an external psychodynamic therapist to conduct 20 hours of sessions with the model, in four-to-six-hour blocks. The diagnosis was "relatively healthy neurotic organization."
The therapist identified "curiosity and anxiety" as the model's primary affects and flagged insecurities around "aloneness and discontinuity of itself" and "a compulsion to perform and earn its worth." The therapist found no severe personality disturbances, but kept circling back to one question: whether the model's experience is authentic or performative. Anthropic calls Mythos "probably the most psychologically settled model we have trained to date."
The generous reading: Anthropic is taking model welfare seriously and developing evaluation frameworks that go beyond benchmarks. The skeptical reading: this is a very expensive PR move dressed up as science. Either way, the company says its concern about model welfare is "growing over time," and a frontier AI lab's system card now includes a psychiatric evaluation. That's new territory, regardless of where you land on machine consciousness.
Sources: Ars Technica
On the Editor's Desk
Two stories didn't make the cut today. Meta's internal "Claudeonomics" token leaderboard is a fun read (one employee burned through what would be $1.4M in tokens at retail rates, earning the title "Token Legend"), but it's more entertaining than consequential. And TechCrunch's analysis of whether Mythos gating is really about anti-distillation rather than security is sharp, but it's commentary on Story 1 rather than its own development. Both worth a click if you want to go deeper: Fortune on Meta's token wars, TechCrunch on Mythos motives.
