“Anthropic used its specialized AI security models to discover approximately 100 potential vulnerabilities in Firefox, which it then responsibly disclosed to Mozilla for remediation.”