JPMorgan's Dimon: Mythos AI Is Like Handing Ballistic Missiles to Individuals
Claire Weston
JPMorgan CEO Jamie Dimon warned at a defense summit that releasing Anthropic's Mythos AI model to the public is like "handing ballistic missiles to individuals" — This means → AI capability has hit the national-security red line, and both government and Wall Street are shifting into defense mode.
What exactly is Dimon afraid of?
Anthropic itself admitted Mythos is too powerful at finding software vulnerabilities to release publicly.
This means → the alarm isn't coming from outsiders — the developer hit the pause button, judging the model's attack potential too dangerous even by its own standards.
Dimon called the threat "real" at the summit; the U.S. government is already paying close attention.
What is JPMorgan doing with its access?
JPMorgan is among a batch of firms that gained Mythos access starting in April.
The bank uses that access to stress-test its own defenses and shares findings with vendors and peers.
In plain terms = the bank isn't using AI to make money — it's using AI to attack itself and find what breaks.
Dimon previously disclosed that hundreds of JPMorgan employees work full-time on hardening systems.
Why does Dimon keep talking about national security?
In his April annual shareholder letter, he wrote that the U.S. needs to "become stronger" to preserve its military and economic power.
JPMorgan announced last year it will invest $1.5 trillion over the next decade in industries that strengthen U.S. economic security and resilience — 50% above its original plan.
This reflects a shift: top Wall Street firms now treat national security as a core business priority, not just a PR talking point.
What does this mean for the AI industry?
The key question: can Mythos be commercially deployed under strict controls?
This means → the outcome will be a landmark test of AI safety governance — release or lockdown will set the ceiling for the entire industry.
In plain terms = if the most powerful models can only stay locked in a lab, AI companies will have to rewrite their business models.
Content is for reference only, not financial advice.