JPMorgan’s Dimon Warns That Anthropic’s New Models Mean Cybersecurity Risks Have Skyrocketed
JPMorgan CEO Jamie Dimon said cybersecurity risks have gone up “10-fold” after Anthropic released its Mythos models.
“AI created vulnerabilities that we didn’t know about, and we always worried about cyber before these things,” he said on Bloomberg TV.
Dimon went on to say that while there are downsides and they are “legitimate,” he wouldn’t get “hysterical” over whether AI poses an existential risk. “What we’re doing is rolling up our sleeves and going to work to fix it,” he said.
The remarks come shortly after OpenAI CEO Sam Altman said society may have to accept harmful effects of AI in exchange for the technology’s broader benefits.
Altman made the comments in an interview with Politico’s Decoded, where he was asked how OpenAI’s approach to AI safety differs from rival Anthropic, which has pushed for greater caution as increasingly powerful models gain the ability to operate more independently.
“I think there’s a lot of daylight,” Altman said of the two companies’ approaches. “One of the differences between us and some of the stricter AI safety people is that we believe that the world should accept some bad things happening for the benefits of this technology and people having the agency,” he added.
Altman argued for what he described as a lighter regulatory approach. “I wouldn’t take a trade of saying we will make sure there’s no major hacks, there’s no misuse of this technology, there’s zero scams,” Altman said, arguing that people would ultimately do “orders of magnitude” more good with AI than harm.
Cybersecurity risks go beyond Anthropic. A report from earlier this year detailed that hackers used a team of autonomous agents to infiltrate Taiwanese government systems, compromise dozens of accounts and steal thousands of personnel records.
The operation, carried out over four days in July, targeted government agencies, critical infrastructure and technology suppliers across Taiwan, according to Taiwanese officials and Israeli cybersecurity company Dream, which discovered the campaign. The AI system mapped 21 government systems, compromised 85 government user accounts and extracted approximately 2,500 personnel records, Dream said.
But what has cybersecurity experts particularly concerned is not simply what the hackers stole. It is how little human involvement may have been required to do it. The attackers assembled an autonomous system capable of coordinating as many as eight AI agents. Those agents performed reconnaissance, attempted to obtain credentials, searched for vulnerabilities, and determined what attack strategies to pursue next, according to Dream.
Rather than serving as a tool controlled step by step by a hacker, the AI effectively operated as its own hacking team.”Like a human team, when an approach gets blocked, it researches new techniques in real time and adapts. It’s an attacker that strategizes, learns, and adjusts on its own,” Amir Becker, Dream’s chief business and strategy officer, told CNN.
The incident marked a potentially significant shift in the rapidly evolving relationship between AI and cybersecurity. Hackers already routinely use generative AI to help write malicious code, research vulnerabilities, or accelerate individual parts of an attack. The Taiwan operation went further by allowing AI agents to coordinate multiple stages of an intrusion and make decisions without continuous human direction.
