AI Apocalypse? Treasury Boss Summons Bank Bigwigs Over Claude Mythos Chaos

by

Well, bugger me with a firewall, the US Treasury’s top dog, Scott Bessent, has rounded up the banking elite for a spot of tea and existential dread. Seems Anthropic’s latest brainchild, the Claude Mythos model, has got everyone’s knickers in a twist over potential cyber shenanigans.

  • Scott Bessent, in a fit of foresight, herded the CEOs of America’s financial titans to discuss the impending doom courtesy of Anthropic’s Claude Mythos AI. All because some clever clogs leaked the code.
  • This Mythos beast has apparently sniffed out more software vulnerabilities than a troll under a bridge finds loose change, sparking fears of hackers having a field day and the financial system going tits up.
  • Meanwhile, Anthropic’s coffers are overflowing with over $30 billion in annualized revenue, thanks to everyone and their dog wanting a piece of the AI pie, including Google and Broadcom. And let’s not forget their Claude Code platform, which is raking it in like there’s no tomorrow.

According to the grapevine, Bessent corralled the banking bigwigs at the Treasury’s swanky Washington digs, with Jerome Powell popping in for a gander. All this fuss kicked off after Anthropic unveiled their Mythos model, which they’ve modestly described as a “cybersecurity nightmare.”

Things went pear-shaped when the model’s code did a runner earlier this month. Anthropic then took to their blog to warn that their AI is better at finding and exploiting software vulnerabilities than a wizard with a spellbook, and the consequences could be as dire as a dragon at a knitting convention.

The meeting conveniently coincided with the banking lot already being in Washington for some industry shindig. Invites were handed out like hotcakes, but only to the heads of banks deemed too big to fail. Because, you know, if they go down, we’re all up the creek without a paddle.

Among the attendees were David Solomon of Goldman Sachs, Brian Moynihan of Bank of America, Jane Fraser of Citigroup, Ted Pick of Morgan Stanley, and Charlie Scharf of Wells Fargo. Jamie Dimon of JPMorgan Chase was invited but apparently had a prior engagement with his cat.

In his annual shareholder letter, Dimon quipped that cybersecurity is still their biggest headache, and AI is just the cherry on top of this anxiety sundae.

Anthropic claims their Mythos model has already found more vulnerabilities than a paranoid spy in a thriller novel. As a result, they’ve locked it down tighter than a troll’s treasure hoard, only letting a select few companies like Amazon, Apple, and Microsoft have a peek.

This is the first time Anthropic has slammed the brakes on a product launch. A few lucky infrastructure and tech firms, such as Cisco and Broadcom, along with the Linux Foundation, have also been granted access. No word yet on whether they’ve had to sign their souls away in the process.

Fears of AI Gone Rogue and Hidden Bugs

Just as everyone was starting to relax, along comes the fear that nefarious types could use AI to crack passwords and encryption like a nut. Anthropic revealed that some of the flaws Mythos uncovered are older than the internet itself, having lain hidden for up to 27 years. Makes you wonder what else is lurking in the digital undergrowth.

This Treasury powwow comes hot on the heels of the US government labeling Anthropic a supply chain risk, a tag the company is fighting in court with all the gusto of a lawyer on double espresso.

Despite all the regulatory side-eye and the Defense Department’s stern looks, Anthropic is laughing all the way to the bank. Their financial momentum is so strong it could power a small country.

In a blog post that dropped on April 6, they boasted that their annualized revenue had rocketed past $30 billion as of early April 2026, up from a mere $9 billion at the end of 2025. That’s what you call a growth spurt.

Much of this success is thanks to their compute partnerships with Google and Broadcom, which are essentially AI’s equivalent of a power-up in a video game. This deal locks in enough TPU capacity to keep their Claude models running until the cows come home.

Their Claude Code platform has been a cash cow, pulling in over $2.5 billion in run-rate revenue by February. Weekly active users have doubled since the start of the year, proving that AI-driven tools are the new black.

Read More

2026-04-10 15:57