Anthropic Teams Up With Its Rivals to Keep AI From Hacking Everything

Anthropic unveiled Mythos Preview and launched Project Glasswing, a cross-industry consortium to assess cyber risks before broad release.

Microsoft, Apple, Google, AWS, the Linux Foundation, Cisco, Nvidia, and Broadcom will get early access to test vulnerabilities and defenses.

The effort reflects growing concern that frontier AI can both discover flaws and help turn them into exploits.

The goal is to use red teaming and coordinated disclosure to strengthen defenses before attackers can benefit.