Anthropic unveiled its latest AI model, Mythos, this month, showcasing advanced features tailored for defensive cybersecurity operations. Concerns arose about Mythos’ extensive capabilities as the AI startup claimed to have identified numerous critical vulnerabilities in major operating systems and web browsers.
The model, dubbed Claude Mythos Preview, was launched as part of “Project Glasswing,” granting access to tech giants such as Amazon, Microsoft, Nvidia, and Apple, along with more than 40 organizations managing critical software infrastructure.
Experts cautioned that Mythos could rapidly detect and exploit unknown vulnerabilities, potentially hastening sophisticated cyberattacks, especially in sectors like banking with intricate and interconnected legacy technology systems. Anthropic warned that if misused, Mythos’ ability to uncover software flaws at scale could pose severe threats to economies, public safety, and national security.
Regulators, including Bank of Canada Governor Tiff Macklem, emphasized the need to address the risks posed by AI advancements like Mythos. Macklem highlighted ongoing discussions among policymakers and financial institutions on handling rapid AI developments and the necessity for robust cybersecurity measures.
The White House engaged in discussions with Anthropic CEO Dario Amodei regarding Mythos, focusing on collaboration, cybersecurity, and balancing innovation with safety. Despite the Pentagon’s supply-chain risk designation on Anthropic, the U.S. government plans to provide a version of Mythos to federal agencies.
Authorities in the U.S., Britain, and Germany expressed concerns about Mythos, initiating dialogues with financial institutions and cybersecurity experts to assess potential risks. Noteworthy critics, like David Sacks, stressed the significance of addressing Anthropic’s AI advancements seriously, acknowledging the potential for increased vulnerability detection and exploit development with more capable coding models.
