JPMorgan $JPM Chase CEO Jamie Dimon on described the risks of broad public access to Anthropic's advanced Mythos AI model as a "real issue," comparing it to handing ballistic missiles to individuals.
"You're giving ballistic missiles to individuals with Mythos, basically," then JPMorgan chief said of the potential for broad release of the Anthropic model
Bloomberg / Getty Images
JPMorgan $JPM Chase CEO Jamie Dimon on described the risks of broad public access to Anthropic's advanced Mythos AI model as a "real issue," comparing it to handing ballistic missiles to individuals.
"You're giving ballistic missiles to individuals with Mythos, basically," Dimon said, in remarks reported by Bloomberg. Speaking at Republican Sen. Dave McCormick's Pennsylvania Defense and Innovation Summit, Dimon added that the U.S. government is on top of the issue now.
The warning centers on Mythos's capacity to identify software vulnerabilities at a level Anthropic has said is too dangerous for general release. JPMorgan is among a select group of companies that gained access to the model in April and has been deploying it to stress-test internal systems and coordinate on findings with vendors and peers. The month after gaining that access, Dimon said the bank had hundreds of employees working full time to strengthen its systems.
JPMorgan's involvement with Mythos is part of its broader push into national security and economic resilience. The bank announced last year a "Security and Resiliency Initiative" through which it would direct $1.5 trillion over ten years toward sectors that strengthen the U.S. economy — a figure 50% above what it would have invested otherwise. JPMorgan also tied the commitment to its separate "American Dream Initiative," pledging $24 million in a mix of loans, investments, and philanthropic grants directed at Philadelphia's shipbuilding sector. Dimon visited the Philadelphia Navy Yard earlier Wednesday as the bank announced that commitment.
The Mythos model has been at the center of a broader government and industry reckoning over AI safety. The U.S. government lifted export controls on Anthropic's Claude Fable 5 and Mythos 5 models this month after an 18-day shutdown triggered by Amazon $AMZN researchers documenting a technique for eliciting dangerous outputs from Fable 5 related to software vulnerability discovery. A limited group of U.S. organizations regained access to Mythos 5 after the government gave its approval on June 26; extending that access to additional domestic and international partners through Anthropic's Glasswing program remains under discussion with federal authorities, Anthropic said.
Dimon has grown increasingly vocal on national security matters in recent years. His April shareholder letter argued that America must "get stronger" if it hopes to preserve its position as a military and economic power.
Join 500,000+ readers who start their day with Quartz.
By subscribing, you agree to our Terms of Service and Privacy Policy.