Anthropic Mythos Finds Financial System Vulnerabilities.

Mazen Al Sabbagh

— 5 min read
Share

TL;DR

  • OpenAI's IPO cleared: The Musk lawsuit was dismissed unanimously and swiftly, removing a significant barrier to OpenAI's public offering.
  • Anthropic validates financial system vulnerabilities: Its Mythos AI briefed the Financial Stability Board on cyber risks, coinciding with urgent IMF warnings.
  • NextEra secures AI power: A record $66.8 billion acquisition of Dominion Energy directly addresses the escalating energy demands of data centers.
  • Vatican and Anthropic ally on AI ethics: Pope Leo XIV's first encyclical, "Magnifica Humanitas," co-launched with Anthropic's Christopher Olah, signals strategic geopolitical positioning.
  • Samsung averts immediate strike: A partial court injunction and direct presidential intervention stabilized market concerns ahead of a critical deadline.

Lead Story: Anthropic Mythos Finds Financial System Vulnerabilities

Anthropic is demonstrating the capabilities of its unreleased Mythos AI model to the Financial Stability Board, revealing cybersecurity vulnerabilities with systemic implications for global financial infrastructure. This briefing, requested by Bank of England Governor Andrew Bailey, directly confronts the emerging risks AI presents to market stability. Reuters reported Monday, citing an FT exclusive.

The FSB, composed of finance ministers and central bank governors from G20 economies, received Anthropic’s insights on Claude Mythos Preview. This cyber-focused model has already identified thousands of critical vulnerabilities across widely deployed software. Through its Project Glasswing program, select institutions, including Apple and JP Morgan, have accessed Mythos, as The Guardian confirmed. An FSB spokesperson noted the body "welcomes engagement with Anthropic and other firms on emerging and frontier risks to global stability."

This engagement arrives as the IMF issued a direct warning that AI has elevated cyber risk to a financial stability threat. Concurrently, the UK’s Bank of England, FCA, and Treasury released a joint statement outlining new expectations for the financial sector regarding frontier AI, cybersecurity, and operational resilience. Bailey’s prior assessment at Columbia University—"You wake up to find that Anthropic may have found a way to crack the whole cyber risk world open"—underscores the urgency.

Anthropic is strategically positioning itself by demonstrating responsible disclosure to the world's leading financial regulators. This move simultaneously reinforces the argument that Mythos is too powerful for broad release, a stance Anthropic has maintained with EU and US authorities. The outcome establishes Anthropic as a primary interlocutor for regulators navigating AI’s profound impact.

In Other News

Musk's OpenAI lawsuit conclusively rejected. A nine-person jury in Oakland dismissed Elon Musk's claims against OpenAI, Sam Altman, Greg Brockman, and Microsoft unanimously within two hours, finding them barred by the statute of limitations. Judge Yvonne Gonzalez Rogers delivered an immediate dismissal, confirming prior readiness to do so. While an appeal is possible, the factual basis of the statute-of-limitations ruling makes reversal challenging. OpenAI’s lead attorney, William Savitt, characterized the verdict as a vindication, a sentiment echoed by Microsoft. This outcome immediately removes the most significant legal impediment to OpenAI's for-profit conversion and its anticipated IPO. The New York Times noted OpenAI's chief strategy officer, Jason Kwon, celebrating the decisive outcome.

NextEra Energy acquires Dominion in historic $66.8 billion power play. Reuters and Bloomberg confirmed NextEra's all-stock acquisition of Dominion, creating the largest regulated electric utility by market value. This transaction explicitly targets the burgeoning AI sector: Dominion supplies northern Virginia’s data center corridor, the world’s largest, with nearly 51 gigawatts of contracted capacity for key tech giants. CNBC highlighted NextEra’s intent to develop over 30 data center hubs nationwide. Dominion shares rose 11%, while NextEra saw a 2.6% decline on dilution concerns. The deal awaits FERC, NRC, and state regulatory approvals, with a projected close in 12-18 months.

Vatican and Anthropic co-founder unveil AI encyclical. The Vatican announced Monday the public release of Pope Leo XIV’s inaugural encyclical, "Magnifica Humanitas," next Sunday. Christopher Olah, Anthropic’s co-founder and a leading interpretability researcher, will speak at the launch. The encyclical, signed May 15th, precisely 135 years after Pope Leo XIII’s "Rerum Novarum" on industrial labor rights, draws a parallel between the Industrial and AI Revolutions in their posing of existential questions. This Vatican alliance represents a notable geopolitical maneuver for Anthropic, particularly amidst its ongoing lawsuit against the Trump administration concerning AI deployment limits.

Samsung shares surge amid strike injunction and presidential intervention. Bloomberg reported a 6% jump in Samsung shares Monday following a court-granted injunction to curb a potential strike and union signals of a willingness to negotiate. South Korea's President Lee Jae-myung personally urged a resolution, escalating from prior prime ministerial intervention. Reuters confirmed that talks resumed, representing the final opportunity before the May 21st walkout of over 45,000 workers.

X / Social Pulse

The Atlantic’s Sunday essay, "AI Has Broken Containment", articulated a critical shift: AI has "ascended to the role of main character" in American discourse, influencing everything from the Trump-Xi summit to mainstream labor anxieties, with nearly three-quarters of employed Americans now anticipating job reductions. Separately, Chamath Palihapitiya's weekend commentary continued to resonate. Beyond his Friendster analogy for Anthropic's safety guardrails, he explicitly criticized PwC and Accenture for "letting the fox in the hen house" by deploying competitor models while AI labs develop their own enterprise solutions. Vanity Fair’s Tom Dotan argued, pre-verdict, that the Musk-Altman trial fundamentally exposed OpenAI’s nonprofit structure as "the most well-funded fig leaf of all time." The jury’s swift legal rejection may have settled one question, but Dotan’s structural critique persists.

One to Watch

Salesforce to spend $300 million on Anthropic tokens while freezing engineer hiring. CEO Marc Benioff disclosed on the All-In podcast Friday that Salesforce anticipates allocating nearly $300 million this year to Anthropic tokens, primarily for AI-powered software development. Benioff credited AI coding tools with over 30% productivity gains, commending Anthropic's performance. Salesforce shares rose nearly 6% on this news. The economic calculus is unambiguous: the investment in AI compute demonstrably offsets human capital costs, and the market is explicitly rewarding this operational shift. This move establishes a clear template for other enterprise CEOs facing similar human capital expenditure decisions.

Quick Hits

  • Nvidia reports Q1 FY27 earnings Tuesday — Morgan Stanley elevated its conviction, predicting a "beat-and-raise" scenario; consensus revenue projections are $78-79B, with market focus on Blackwell ramp and forward guidance.
  • OpenAI launches personal finance in ChatGPT via Plaid — Pro users can now integrate bank accounts from over 12,000 institutions to access spending analytics, portfolio tracking, and personalized financial guidance powered by GPT-5.5; read-only access is in place, with Intuit integration slated for future release.
  • Colorado repeals and replaces its AI Act — Governor Polis signed SB 189 on May 14, pivoting from a broad "high-risk AI system" framework to more focused regulations on automated decision-making; Georgia simultaneously enacted a targeted chatbot disclosure law (SB 540).
  • Arm Holdings faces FTC antitrust probe over chip licensing practices following the launch of its own AGI CPU — Bloomberg/Reuters indicate the FTC is investigating potential restrictions on architecture access for rivals; ARM shares dropped 8.5% Friday.
  • Europe's electricity costs threaten its AI ambitions — CNBC reported that U.K. power costs are $111.65/MW compared to $28/MW in the U.S.; data center capacity costs in Europe's five largest markets are projected to increase 12% in 2026.

Monday delivered a verdict in more ways than one. The Musk-OpenAI jury took 90 minutes to decisively close a chapter Silicon Valley had debated for two years—the statute of limitations argument dismantled the last legal avenue to force OpenAI’s structural rollback. This clears the runway for its IPO. Concurrently, Anthropic cultivates a different kind of market advantage, building regulatory relationships through the direct demonstration of Mythos to the FSB. The NextEra-Dominion megadeal quantifies the market’s re-evaluation of energy infrastructure, valuing the AI-driven power demand thesis at $66.8 billion. Tomorrow’s Google I/O keynote will further test Gemini’s competitive standing at a fraction of the cost. The through-line is unmistakable: the foundational architecture governing capital, processing, energy, and governance is undergoing a simultaneous, AI-driven transformation.

Sources

Lock in. M. mazen@thorterminal.com

Read more