Select Language:
Recent reports reveal that the U.S. Treasury Department and the Federal Reserve have recently pressed several major banks to experiment with Anthropic’s latest AI model, Mythos. This model is specifically designed to identify potential vulnerabilities within financial systems, highlighting its importance for cybersecurity in critical infrastructure.
Anthropic officially unveiled the Mythos model on April 7, 2026, but immediately placed restrictions on its accessibility. Currently, only a select handful of companies—including Amazon, Apple, Broadcom, Cisco, Google, Microsoft, NVIDIA, and JPMorgan Chase—have been granted initial access under a project code-named Glasswing. The project aims to bolster security for essential software infrastructure by leveraging advanced AI capabilities.
While Anthropic clarifies that Mythos was not developed explicitly for cybersecurity training, the model has demonstrated remarkable proficiency in discovering system flaws. Industry analysts suggest that the limited release could also serve as a strategic marketing move, creating a sense of scarcity that boosts the product’s appeal.
On April 7, key bank executives were summoned to the Treasury Department’s headquarters, where they were warned about the critical importance of deploying Mythos’s vulnerability detection features. The meeting involved staff from systemically important banks, including JPMorgan Chase, Goldman Sachs, Citigroup, Bank of America, and Morgan Stanley. Authorities emphasized the necessity of taking the model seriously given its potential to reinforce financial security.
In the meantime, Goldman Sachs, Citigroup, Bank of America, and Morgan Stanley are reportedly conducting internal tests of Mythos and expect to receive full access within a few days. U.S. National Economic Council Director Lael Brainard commented that Treasury Secretary Janet Yellen’s approach was appropriate amid a climate of urgency. She also noted that Anthropic has agreed to withhold the model from public release until regulators fully understand its implications.
Interestingly, this cautious stance contrasts sharply with earlier conflicts involving Anthropic. In February, the Department of Defense designated Anthropic a supply chain risk, prompting former President Donald Trump to demand federal agencies cease using the company’s technology. In March, Anthropic filed a lawsuit against the Trump administration, alleging unprecedented and unlawful actions. A federal judge in California later issued a temporary injunction, blocking government efforts to restrict the AI company’s operations.
Meanwhile, data from Ramp, an enterprise expenditure platform, indicates that corporate AI adoption hit a milestone in March, surpassing 50% for the first time at 50.4%. Notably, Anthropic’s market share grew from 24.4% to 30.6%, narrowing the gap with OpenAI to just 4.6 percentage points. Ramp predicts that Anthropic could overtake OpenAI within the next two months, signaling a rapidly shifting landscape in AI utilization among major corporations.
ChatGPT
Perplexity
Gemini AI
Grok AI
