Catenaa, Tuesday, May 12, 2026- Anthropic Chief Executive Dario Amodei warned that artificial intelligence is exposing software vulnerabilities faster than they can be fixed, creating a narrow window to prevent widespread cyberattacks, during a discussion last week with JPMorgan Chase CEO Jamie Dimon.
Amodei said the company’s AI systems have uncovered tens of thousands of software flaws across widely used platforms. He warned that organizations may have only six to 12 months to patch critical weaknesses before similar tools become broadly accessible to attackers.
He said the growing speed of AI-driven vulnerability discovery could increase risks of ransomware attacks, data breaches, and financial disruption across sectors including healthcare, education, and banking.
Anthropic’s internal testing showed its model identifying hundreds of vulnerabilities in a single scan of major software systems, highlighting how AI can process complex codebases at scale far faster than human researchers.
The company has restricted access to its most advanced security tools through a limited program with selected partners, aiming to support patching efforts before broader release. Amodei said some vulnerabilities remain undisclosed until fixes are developed to avoid exploitation.
The warnings come amid debate within the AI industry over the scale of the risk. OpenAI Chief Executive Sam Altman has suggested some concerns may be overstated, arguing that competitive positioning and safety messaging are both shaping public discussion around AI security.
Despite the debate, reports indicate government agencies are already using Anthropic’s systems in cybersecurity applications to scan networks for weaknesses.
Amodei said the current period represents a critical phase in which rapid response by companies and governments could determine whether AI-driven cyber risks are contained or escalate further.
He said the number of software bugs is finite, but the speed at which they are discovered has changed the security landscape.
