An artificial intelligence startup has unveiled a new cybersecurity collaboration with some of the world’s largest technology companies on Tuesday, according to a company announcement.
Anthropic revealed its “Project Glasswing” initiative, which grants selected organizations early access to test an unreleased AI system called “Claude Mythos Preview” specifically designed for cybersecurity defense purposes. The partnership includes tech giants Amazon, Microsoft, Apple, Google, and Nvidia, along with cybersecurity specialists CrowdStrike and Palo Alto Networks.
The launch comes after reports last month indicated Anthropic had been testing Claude Mythos, which reportedly demonstrated both significant security risks and advanced defensive capabilities. Those reports had caused stock prices for cybersecurity companies like Palo Alto Networks and CrowdStrike to drop substantially.
Recent discussions at this year’s RSA cybersecurity conference in San Francisco centered heavily on the growing threat of artificial intelligence-powered cyber attacks and questions about whether traditional security measures remain adequate.
According to Anthropic’s Tuesday blog post, the Mythos Preview system has already identified “thousands” of significant security flaws across operating systems, web browsers, and various software applications.
The company stated that participating organizations will integrate Mythos Preview into their defensive security operations, while Anthropic plans to distribute the findings throughout the cybersecurity industry.
Beyond the initial partners, Anthropic announced it will expand access to approximately 40 additional organizations that manage essential software infrastructure. The company has committed up to $100 million in usage credits and $4 million in donations to open-source security organizations.
The AI company expressed its ultimate objective as enabling “our users to safely deploy Mythos-class models at scale.”
Anthropic confirmed it has maintained ongoing communications with the U.S. government regarding the model’s capabilities.
The initiative follows previous security incidents, including last year’s attack where hackers exploited weaknesses in Anthropic’s Claude AI system to target approximately 30 organizations worldwide. Additionally, a recent study conducted by IBM and Palo Alto Networks found that 67% of 1,000 surveyed executives reported experiencing AI-based attacks in the past year.
