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

image via WIRED

April 7, 2026, 6:49 PM

• •Mythos Preview can identify vulnerabilities and develop exploits.
• •Project Glasswing aims to address cybersecurity implications of advanced AI.
• •The consortium includes major tech companies.
• •Anthropic is prioritizing coordinated vulnerability disclosure practices.
• •The project seeks to prepare for a future with widely available powerful AI models.

Anthropic unveiled its new AI model, Mythos Preview, and launched Project Glasswing, an industry consortium, to tackle the cybersecurity challenges posed by advanced AI capabilities. The consortium, including tech giants like Microsoft, Apple, and Google, will have private access to the model to mitigate vulnerabilities before its wider release. Mythos Preview excels at finding vulnerabilities, including those decades old, and developing exploits, thus potentially upending current software security practices. The initiative emphasizes the need for urgent collaboration across the industry to prepare for a future where powerful AI models are broadly accessible.

Read original article