Anthropic Says Its New AI Model Is So Good at Finding Security Risks, You Can't Use It

Anthropic has unveiled its latest AI model, Claude Mythos, which excels at identifying cybersecurity vulnerabilities to such an extent that it cannot be released publicly.

Instead, the model will be made available to major tech companies as part of a new cybersecurity consortium called Project Glasswing, which includes industry giants such as Apple and Microsoft.

The model has reportedly discovered thousands of severe security flaws across various operating systems and web browsers. Anthropic's CEO, Dario Amodei, emphasized the potential to create a more secure internet if the consortium effectively addresses the risks posed by AI capabilities.

The company is investing $100 million in usage credits for the model and donating $4 million to open-source security organizations, highlighting its commitment to improving cybersecurity in the face of advanced AI threats.