Anthropic’s “Mythos” AI Sparks Cybersecurity Fears as Company Limits Release
Anthropic’s “Mythos” AI Sparks Cybersecurity Fears as Company Limits Release
———————————
Artificial intelligence company Anthropic has developed a new system reportedly capable of identifying software vulnerabilities across major operating systems and web browsers, raising concerns among cybersecurity experts and policymakers about potential misuse.
According to company disclosures, the tool—referred to as “Mythos”—has already identified thousands of security weaknesses and could be used to strengthen digital infrastructure. However, Anthropic has not released the model publicly, citing risks that it could be exploited by malicious actors.
Instead, the company is reportedly sharing access under a restricted program known as “Project Glasswing” with selected major technology and financial firms, including Amazon, Apple, Cisco, JPMorgan Chase, and Nvidia, to help them test and improve their cybersecurity defenses.
The development has drawn attention from government and financial leaders. Senior U.S. officials and banking executives have reportedly discussed AI-related cyber risks in closed-door meetings, reflecting growing concern that advanced AI systems could accelerate cyberattacks on critical infrastructure.
Security experts warn that AI tools are already being used to enhance phishing campaigns, malware development, and automated cyberattacks, reducing the time needed for hackers to exploit vulnerabilities. Analysts say AI significantly increases both the speed and scale of cyber threats compared to human-led operations.
Critics have also questioned whether Anthropic’s controlled rollout is motivated partly by commercial strategy, especially as the company is widely reported to be considering future public listing plans. Others argue the restricted release aligns with its stated focus on AI safety and responsible deployment.
Anthropic has emphasized that misuse of such technology could pose serious risks to economies, public safety, and national security, and says its cautious approach is intended to prevent harmful deployment while still enabling defensive cybersecurity improvements.
