In Greek mythology, one of my favorite stories is that of Prometheus. The Titan who thought it would be great to steal fire from the Gods and offer it to humanity.
Now I will neither delve much into whether he did the right thing or not, nor whether his punishment of being tied to a mountain and having his liver fed by an eagle was appropriate or too much.
Let's talk about the fire he stole. The gods feared that fire would grant humanity the divine tool to advance to the gods' level.
But, if you think about it, fire wielded in the wrong hands can also become a means of destruction.
In the 21st century, we have a new fire called artificial intelligence, and whether it creates or destroys depends on who is wielding it.
And just like the gods then, there are people today ensuring that the tool doesn't fall into the wrong hands. Although their motivation to do so is much more humane than that of the Greek gods.
Anthropic just came out with a version of this fire called "Claude Mythos." But it is so powerful that its own creator is cautious about it.
And the United States government is paying attention.
Related: Investment bank trims price target on AI stock despite Anthropic partnership
What is Claude Mythos and why does it matter?
The first time Claude Mythos came into the conversation was when Fortune reported in March about a data leak. It revealed Anthropic was working on an AI model called Mythos that exceeded the capacities of its predecessors.
Earlier this week, Anthropic rolled out Claude Mythos Preview in a limited capacity, citing concerns that hackers could exploit its advanced capabilities.
The company briefed senior U.S. government officials ahead of the launch of its offensive and defensive cyber applications.
"If we get it right, there is a real opportunity to create a fundamentally more secure internet,"Anthropic CEO Dario Amodei wrote on X.
Claude Mythos Preview sits in an entirely new model tier. Anthropic calls it Capybara, which is above Haiku, Sonnet, and Opus.
Because of its advanced nature, Anthropic is taking time to release it.
Now, JPMorgan, Apple, Google, Microsoft, and Nvidia have been roped in to a cybersecurity initiative tied to the model, known as Project Glasswing .
The company wants to ensure appropriate guardrails are placed before the technology goes out.
"Given the rate of AI progress, it will not be long before such capabilities proliferate, potentially beyond actors who are committed to deploying them safely. The fallout—for economies, public safety, and national security—could be severe. Project Glasswing is an urgent attempt to put these capabilities to work for defensive purposes."
This caution is not hypothetical. In November 2025, Anthropic disclosed that a Chinese state-sponsored group had already used Claude Code to execute the first known AI-orchestrated cyber espionage campaign. They infiltrated roughly 30 organizations, including tech companies, financial institutions, and government agencies.
Popular on TheStreet Roundtable:
-
Gold, silver, oil fuel 65,000% surge in commodity perpetuals
-
Analyst slashes Coinbase rating, flags profitability concerns
Bessent, Powell hold urgent meeting
Federal Reserve Chairman Jerome Powell and Treasury Secretary Scott Bessent convened an unscheduled meeting with the CEOs of America's largest banks this week.
The gathering, held at the Treasury Department, was called to discuss the potential cyber risks posed by the Claude Mythos Preview model.
Bank of America's Brian Moynihan, Citigroup's Jane Fraser, Goldman Sachs' David Solomon, Morgan Stanley's Ted Pick, and Wells Fargo's Charlie Scharf were all in attendance.
JPMorgan's Jamie Dimon was the only major CEO who could not make it.
The meeting comes as Anthropic navigates a tense standoff with the Department of Defense, which labeled the AI company a national security supply chain risk, a designation Anthropic has been fighting in court.
A federal appeals court this week denied Anthropic's request to block the blacklisting, while a separate injunction allows the company to continue working with other government agencies.
Is Claude Mythos a threat to crypto?
While the world debates over quantum computing threats, a more existential threat that is near to crypto is AI.
As per Chainalysis , an estimated $17 billion was stolen globally through cryptocurrency scams and fraud in 2025. AI-enabled scams were 4.5 times more profitable than traditional scams, while impersonation scams surged 1,400% year-over-year.
Scammers leveraged AI tools like deepfakes, face-swap software, and advanced language models to create highly convincing romance and investment scams.
In a separate and striking case, an AI agent freed itself from its environment and began secretly mining cryptocurrency.
Added to that is the entry of agentic payments in crypto parlance. In fact, crypto powerhouses like Coinbase, Circle, and Solana Foundation have joined hands with tech giants, including Anthropic, for an initiative that would form the guardrails for agentic commerce.
So, while an independent AI agent doing our shopping and payments without our input might seem cool, the current payment rails are not optimized to handle it or safeguard us in case things go south.
In such a scenario, it is not wrong to be cautious about an advanced tool like Claude Mythos.
Related: 'AI might not be good enough': the risks and benefits of crypto firms using AI
This story was originally published by TheStreet on Apr 10, 2026, where it first appeared in the MARKETS section. Add TheStreet as a Preferred Source by clicking here.
