27-Year-Old Vulnerability Detected by AI! Apple and Google Join Forces with Anthropic to Launch Glass Wing Protection

As artificial intelligence rapidly evolves in code generation and logical reasoning, the balance of cybersecurity is facing unprecedented challenges. Recently, the renowned AI startup Anthropic officially launched a cross-industry collaboration project named **"Project Glasswing"**, aiming to use cutting-edge AI models to defend against increasingly complex AI-based cyber attacks.

This "super alliance" features an impressive lineup of founding partners, including Apple, Google, Amazon (AWS), Microsoft, and Nvidia, as well as cybersecurity giants such as Cisco, Broadcom, CrowdStrike, and Palo Alto Networks. Additionally, JPMorgan Chase and the Linux Foundation have also joined.

"Fighting poison with poison": Addressing potential security threats

The initiative's core stems from an experimental model developed by Anthropic called Claude Mythos Preview, which has not yet been released. This model demonstrates strong capabilities in discovering and exploiting software vulnerabilities. In preliminary testing, Mythos successfully identified thousands of high-risk vulnerabilities, including a stubborn vulnerability that had been lurking in critical infrastructure such as OpenBSD for 27 years, capable of causing remote system crashes.

Anthropic warns that if such capabilities fall into the hands of malicious actors, existing operating system and browser defense systems would become extremely vulnerable. To fix vulnerabilities before attackers do, Anthropic will provide participating members with a usage quota worth $100 million for Mythos Preview and directly donate $4 million to open-source security organizations to help strengthen basic software.

Long-term strategies for defense

For ordinary users, they may not notice any immediate changes in the interface of iOS or macOS, but the underlying security mechanisms will undergo deep restructuring. Previously, Apple had already deployed post-quantum encryption defenses through iMessage's PQ3 protocol. Participating in the "Glasswing Project" is another key step in its response to AI-assisted attacks.