Anthropic CEO Dario Amodei warned that artificial intelligence has revealed a critical window for major institutions to patch tens of thousands of software vulnerabilities. His company's advanced model, Mythos, uncovered deep-seated flaws across critical systems used by tech firms, governments, and banks.
The Threat Revealed by Mythos
Amodei revealed that Anthropic's model, Mythos, has unearthed numerous vulnerabilities, some of which are decades old, within essential software infrastructure. The scale of these potential weaknesses is significant:
- An earlier Anthropic model identified approximately 20 vulnerabilities in the Firefox browser.
- Mythos alone found nearly 300 vulnerabilities.
- The cumulative total across all tested software now reaches the tens of thousands.
Timeline for Remediation
According to Amodei, the time available to fix these issues is estimated to be between six to twelve months. This timeline is partly based on the current development pace of AI models from geopolitical rivals, which he suggested might lag behind Anthropic's offerings.
Potential Impact and Industry Concerns
Speaking alongside JPMorgan Chase CEO Jamie Dimon, Amodei highlighted the potential financial fallout from cyberattacks. He warned of an "enormous increase in the amount of vulnerabilities, in the amount of breaches, in the financial damage that's done from ransomware on schools, hospitals, not to mention banks."
Due to the high risk of exploitation by malicious actors or adversarial nations, Anthropic has restricted access to the Mythos model, limiting it to select partner companies.
Unpatched Flaws and Exploitation Risk
Amodei noted that the majority of the vulnerabilities discovered by Mythos have not been publicly disclosed because they remain unpatched. He cautioned that if these flaws are identified by malicious parties, they are highly likely to be exploited.
