Anthropic investigating possible breach of its Mythos AI model
CBS News • • business
Key Points:
- Anthropic is investigating a potential security breach involving Mythos, an AI model designed to detect software vulnerabilities, after reports of unauthorized access in a third-party vendor environment.
- The company has not found any breaches beyond the vendor environment or any direct compromises to its own systems.
- Mythos was released in April to a limited group of major companies, including Amazon, Apple, Cisco, JPMorgan Chase, and Nvidia, as part of Project Glasswing to enhance cybersecurity defenses.
- Despite its intended use to improve security, experts and officials have expressed concerns that Mythos could be exploited by malicious actors to target critical IT infrastructure in sectors like banking, healthcare, and government.
- Anthropic confirmed the investigation following Bloomberg's report of unauthorized users accessing Mythos, highlighting the delicate balance between advancing AI cybersecurity tools and managing associated risks.
