Anthropic CEO and co-founder Dario Amodei (left) and JPMorgan Chase CEO Jamie Dimon.
Dennis Bariboos | Reuters | Samuel Corum | Bloomberg | Getty Images
Anthropic CEO Dario Amodei warned on Tuesday that artificial intelligence has narrowed the window for the world’s tech companies, governments and banks to fix tens of thousands of software vulnerabilities found in the company’s latest models.
Its AI model, Mythos, was previewed last month and revealed to have discovered decades-old vulnerabilities in critical software.
Geopolitical adversary China’s AI models are “probably 6 to 12 months” behind Anthropic’s products, so Amodei said it has “about the same amount of time” to resolve these issues.
The comment was made during an Anthropic event where Amodei shared the stage. JP Morgan Chase CEO Jamie Dimon announced a new suite of agents aimed at automating financial operations.
“The danger is that the amount of vulnerabilities, breaches and economic damage caused by ransomware for schools and hospitals, as well as banks, has increased significantly,” Amodei said.
Anthropic limits Mythos to a small number of partner companies due to concerns about what criminals and hostile nations might do with Mythos. The company’s past few model updates have resonated in the market, but Mythos is causing the most concern from both businesses and policymakers.
Amodei said the scale of a potential cyberattack grows with each generation of Claude. Previous Anthropic models discovered approximately 20 vulnerabilities in the Firefox browser. Mythos has discovered nearly 300, and the total number of all software is now in the tens of thousands, he said.
Most of the vulnerabilities discovered by Mythos are not disclosed because they are not patched, and if discovered, “bad actors would exploit them,” Amodei said.
“A better world”
Despite the caution, Amodei and Dimon also expressed some conditional optimism.
“This is a moment of crisis, and I think if we respond to it correctly, if we start taking the first steps, we will have a better world on the other side,” Amodei said. “There are only so many bugs you can find.”
Dimon also said that while cyber concerns are justified, the cybersecurity risks created by AI are “in transition.”
On regulatory issues, Amodei said AI surveillance is similar to what is being done in the auto industry and needs to balance consumer safety with industry competition.
“You can’t start a car company without asking, ‘Does this thing have brakes?'” he says. “We need to look for some process that allows the industry to operate quickly, that is fair, but that puts guardrails around the most serious matters.”
The company’s event and its setting with Dimon, the finance industry’s best-known spokesperson, appeared to demonstrate Anthropic’s advantage over OpenAI in the enterprise AI market as both companies move toward potential IPOs.
Anthropic on Tuesday announced enhancements to its financial services platform, including 10 new AI agents for investment banking and back-office operations, as well as integrations between various Microsoft Office programs. The company also said its latest widely available model, Claude Opus 4.7, leads the benchmark for financial analysis tasks.
