U.S. AI Firm Grants India Access To Cybersecurity Model

Anthropic, a U.S.-based AI firm, has extended access to its Claude Mythos Preview model to India as part of Project Glasswing, its cybersecurity-focused initiative. The program now includes about 150 organizations across 15 countries, including France, Germany, Japan, and Canada, all operating critical infrastructure relied upon by millions. Claude Mythos Preview is tailored for advanced cybersecurity research, including code analysis and vulnerability detection. Anthropic states the model can uncover security flaws in large software systems that might otherwise remain hidden. Access remains restricted to select partners through Project Glasswing, with no public release planned. Launched in April 2026, Project Glasswing initially served around 50 partners, including tech firms, financial institutions, and open-source organizations. The latest expansion marks its first significant rollout beyond the U.S. and U.K. Anthropic emphasizes that participating organizations share a high-risk profile, where a major cyberattack could impact over 100 million people. The company reports that organizations using the model have already identified over 10,000 high- or critical-severity software vulnerabilities. Anthropic highlights the global security implications, noting that successful attacks on these systems could disrupt national and international stability. India joins a select group of countries granted access, though the specific Indian organizations involved have not been disclosed. The initiative underscores Anthropic’s focus on collaborative cybersecurity innovation to protect critical digital infrastructure worldwide.