Anthropic is limiting access to a new AI system, Mythos, after the model reportedly surfaced thousands of vulnerabilities across major operating systems and web browsers, raising fears of rapid weaponization by hackers. The company launched “Project Glasswing” to give select partners including Amazon, Apple, Cisco, JPMorgan Chase and Nvidia early access to harden defenses before wider release. U.S. officials, including the Treasury secretary and Federal Reserve chair, have engaged bank CEOs on AI-driven cyber risks, while the IMF warned that the global financial system remains ill-prepared for massive cyber threats. Security experts say adversaries already use AI to scale phishing, malware and ransomware, compressing attack timelines. Some analysts also question whether the cautious rollout doubles as marketing ahead of anticipated IPOs in the sector, even as Anthropic underscores its safety-first stance.
Related articles:
NIST AI Risk Management Framework (AI RMF 1.0)
MITRE ATLAS: Adversarial Threat Landscape for AI Systems
OWASP Top 10 for LLM Applications
