Mythos Triggers Emergency Treasury Meeting as Cyber Stocks Collapse
Less than 48 hours after Anthropic unveiled Project Glasswing, the fallout has reached the highest levels of the US financial system. Treasury Secretary Scott Bessent and Federal Reserve Chair Jerome Powell convened an urgent closed-door meeting with the CEOs of America’s largest banks this week, warning them about the cybersecurity threats posed by Anthropic’s Mythos AI model.
🏛️ The Emergency Meeting
The meeting, held Tuesday at the Treasury Department in Washington, D.C., was not on any official schedule. It was called specifically to address what Mythos — an AI model that found thousands of previously unknown vulnerabilities in every major operating system and web browser — could mean for the financial system’s resilience.
Bloomberg first reported that the CEOs of Citigroup, Morgan Stanley, Bank of America, Wells Fargo, and Goldman Sachs were among those present. JPMorgan’s Jamie Dimon was notably absent, according to one source.
The timing was strategic: many of the bank executives were already in Washington for separate industry events. The Treasury and Fed used that coincidence to convene what amounted to a national security briefing on AI-driven cyber risk.
Banking regulators and financial officials used the gathering to stress proactive defense strategies, urging institutions to assess and strengthen their cybersecurity infrastructure immediately. The message was clear: the threat landscape has fundamentally changed, and current defenses may not be enough.
📉 The Market Collapse
The financial markets heard the message too — and panicked.
Cybersecurity stocks were among the worst performers in the S&P 500 on Friday. CrowdStrike dropped 6%, Palo Alto Networks fell 6%, Zscaler dropped 6%, and Okta tumbled 7%. The selloff was triggered by the realization that an AI model capable of finding vulnerabilities faster than any human could fundamentally disrupt the traditional cybersecurity business model.
Bernstein analyst Peter Weed noted that investors were selling “like clockwork” — a pattern that has repeated every time Anthropic releases a new cybersecurity capability. But Weed argued the selloff may be overblown: “This does not appear to reduce the potential cybersecurity sector tailwinds” from generative AI.
D.A. Davidson analyst Rudy Kessinger acknowledged that “headline risk” will continue to create volatility, but added that it’s “highly unlikely that tools from the frontier labs will displace cybersecurity vendors, particularly those that primarily provide solutions that do real-time detection & response.”
The counter-argument is straightforward: AI doesn’t eliminate the need for cybersecurity. If anything, it increases it. More powerful AI means more powerful attackers, which means more demand for defense. The question is whether today’s cybersecurity companies are the ones providing that defense, or whether AI-native approaches make them obsolete.
⚖️ The Regulatory Dilemma
The Treasury meeting exposes a deep tension at the heart of AI regulation.
Anthropic briefed senior US government officials about Mythos’s capabilities before the public announcement. It restricted access to approximately 40 partner organizations — including Apple, Google, Microsoft, and Amazon — specifically to prevent offensive use. It committed $100 million in computing resources for defensive security work.
But the genie is out of the bottle. As Anthropic CEO Dario Amodei acknowledged: “More powerful models are going to come from us and from others, and so we do need a plan to respond to this.”
The government faces an impossible choice: regulate too aggressively and push AI development underground or to adversarial nations; regulate too loosely and leave critical infrastructure exposed. The Treasury meeting suggests officials are choosing the middle path — briefing, warning, and urging voluntary action rather than imposing restrictions.
🔐 What the Banks Should Be Worried About
The banking system is uniquely vulnerable to Mythos-class AI for three reasons:
-
Legacy infrastructure. Banks run on systems that are decades old. The 27-year-old OpenBSD vulnerability that Mythos found isn’t an outlier — it’s the norm. Every major financial institution has software dependencies with unknown vulnerabilities.
-
Speed of exploitation. CrowdStrike’s CTO put it bluntly: “The window between a vulnerability being discovered and being exploited by an adversary has collapsed — what once took months now happens in minutes with AI.” Banks that took weeks to patch critical vulnerabilities now have minutes.
-
Systemic interconnectedness. A vulnerability in one bank’s systems can cascade through interbank payment networks, clearing houses, and settlement systems. The Treasury and Fed aren’t worried about one bank being hacked — they’re worried about the entire system being compromised simultaneously.
🏦 The Anthropic Valuation Story
While cybersecurity stocks were tanking, Anthropic’s own trajectory went the other direction.
On Ventuals — a platform that lets traders speculate on private company valuations — Anthropic’s implied valuation briefly overtook OpenAI’s this week, with Anthropic trading at $846 billion against OpenAI’s $838.95 billion. That’s a stunning inversion for a company valued at $61.5 billion just over a year ago.
To be clear: this is a speculative market signal, not an official valuation. Anthropic’s last funding round in February valued it at $380 billion. But the direction of travel is unmistakable: Anthropic’s revenue hit a $30 billion annualized run-rate in early April, up from $9 billion at the close of 2025 — a 3x increase in roughly three months.
🔍 The Bottom Line
This is what the AI security era looks like. It’s not a theoretical future — it’s happening right now, in real time, at the highest levels of government and finance.
The emergency Treasury meeting is the signal. When the US Treasury Secretary and the Fed Chair call an emergency meeting with Wall Street CEOs about an AI model, that model has crossed from a technology story into a national security story.
The market crash is the reaction. Investors aren’t wrong to be scared — they’re just scared about the wrong thing. The real threat isn’t that AI replaces cybersecurity companies. It’s that AI makes every existing vulnerability exponentially more dangerous.
The Anthropic story is the irony. The same company that triggered the panic is also the one leading the defensive response. Project Glasswing is a genuine attempt to use Mythos for good. But as Amodei himself acknowledged, adversaries will inevitably develop similar capabilities — and they won’t restrict access to defenders.
What changes this week: Banks will accelerate cybersecurity spending. Regulators will push for faster patching requirements. Cybersecurity companies that adapt to AI-native defense will thrive. Those that don’t will become cautionary tales.
What doesn’t change: The fundamental asymmetry of cybersecurity — attackers only need to find one vulnerability; defenders need to find all of them. AI just made that asymmetry much, much worse.
Sources
- Bloomberg: Anthropic’s Mythos sparks Washington’s big bank anxiety
- EconoTimes: Anthropic’s Mythos AI Model Sparks Emergency Cybersecurity Meeting With Top US Bank CEOs
- Morningstar/MarketWatch: Palo Alto Networks and other cybersecurity stocks slide on fresh Anthropic fears
- AFP/TechXplore: After Anthropic’s Mythos AI uncovers thousands of zero-day bugs, top US officials huddle with bank CEOs
- OfficeChai: Anthropic Surpasses OpenAI In Valuation On A Private Market Valuation Platform
