Anthropic collaborates with Amazon, Microsoft, and Apple to test an advanced AI model in cybersecurity. The project includes partners such as CrowdStrike and Google.

The artificial intelligence company Anthropic has announced a major initiative 'Project Glasswing', in collaboration with major technology companies such as Amazon, Microsoft, and Apple. This initiative will allow selected partners to test an advanced AI model, called 'Claude Mythos Preview', intended for defensive activities in the field of cybersecurity. Among the partners are also CrowdStrike, Palo Alto Networks, Google, and Nvidia. The announcement comes after Fortune magazine reported that the Claude Mythos model, with advanced capabilities, also raises security risks, which has led to declines in the stocks of some companies in the cybersecurity sector. The Mythos Preview model has identified 'thousands' of vulnerabilities in operating systems and software, and partners will use these findings for defensive activities. Anthropic will expand access to about 40 organizations responsible for critical software infrastructure and will provide usage credits of up to 100 million dollars, as well as donations of 4 million dollars for open-source projects. The startup is in discussions with the U.S. government regarding the model's capabilities, considering that last year hackers exploited vulnerabilities in the Claude system to attack 30 global organizations. A recent study shows that 67% of interviewed executives were targeted by AI-based attacks in the past year.