TOI correspondent from Washington: In the rapidly escalating global race to harness artificial intelligence for cyberwarfare and cyberdefense, India finds itself in an unusually exclusive club.Anthropic, the US artificial intelligence company behind Claude AI, this week expanded access to its highly restricted cybersecurity model, Mythos, to roughly 150 organizations across more than 15 countries, including India, under an initiative called Project Glasswing. India’s inclusion is notable because the list is otherwise dominated by Washington’s closest security partners and allies, with China a conspicuously absentee.Project Glasswing is Anthropic’s attempt to address what it believes is a looming cybersecurity crisis. The company says Mythos has reached a level of coding and vulnerability-discovery capability that surpasses almost all human security researchers. In testing, the model reportedly uncovered thousands of serious software vulnerabilities across major operating systems and browsers, helping partners identify and patch flaws before malicious actors can exploit them.Anthropic has been unusually cautious with Mythos. Unlike consumer AI products released to millions of users, Mythos remains gated because the same capability that can find security weaknesses can also potentially be used to exploit them. British and American security researchers have warned that frontier AI systems are approaching the point where they could significantly accelerate sophisticated cyberattacks.That explains why access is limited largely to governments, critical infrastructure operators, major technology companies, financial institutions and cybersecurity agencies. Participants are expected not merely to use Mythos, but to help test it, identify risks and improve defenses.For India, the invitation carries strategic significance. New Delhi is not a treaty ally of Washington, yet it is the only major non-allied power included in the current expansion. Other countries that have access to Mythos include France, Germany, Italy, Switzerland, the Netherlands, Spain, Belgium, Sweden, Canada, Japan, South Korea, Australia, and New Zealand.The decision to include India likely reflects a combination of geopolitical and commercial realities. First, India has become one of the world’s largest pools of software talent and one of the fastest-growing markets for AI adoption. Second, India’s digital infrastructure, from banking and payments to telecom and public digital platforms, has become so extensive that vulnerabilities discovered in Indian systems can provide valuable insights into securing software globally.This has fueled a more cynical interpretation among some technology observers, one of whom posted on X: Anthropic is not doing India a favor by inviting it to lunch; India is the lunch. The argument is that India’s enormous software ecosystem offers a uniquely rich environment for stress-testing frontier cybersecurity AI. By allowing Indian institutions into Project Glasswing, Anthropic gains access to a vast landscape of real-world code, infrastructure and security challenges. Others believe though that the relationship will be mutually beneficial.Whether India is the “largest ungated AI market in the world” is harder to prove definitively. While India is certainly among the fastest-growing AI user bases globally, per-capita adoption remains lower than in wealthier countries. Still, for American AI companies seeking scale, talent and enterprise customers, India is hard to ignore.China’s exclusion is less mysterious. Project Glasswing is fundamentally a security initiative, and Mythos is designed to discover software vulnerabilities at scale, making it a technology with obvious national-security implications. At a time of intensifying US-China technological rivalry, U.S lawmakers like Senator Chris Coons ensured there was no chance Washington would allow such capabilities to Chinese institutions.India also enjoys another advantage inside Anthropic itself. The company has a significant Indian presence in its leadership and engineering ranks, including Chief Technology Officer Rahul Patil, formerly CTO of Stripe. Patil is among a growing number of Indian-origin executives helping shape the next generation of frontier AI infrastructure.Anthropic has also opened a global office in Bengaluru, only the second after Tokyo, tapping local engineering talent to build what it describes as “sovereign” and localised AI infrastructure. Last October, CEO Dario Amodei met PM Narendra Modi, who is all in on tech, to discuss deploying AI across education, healthcare and agriculture, while noting that usage of Claude Code in India had risen five-fold in recent months.For Indian banks and critical infrastructure operators, the development could ultimately be positive. If Mythos can identify vulnerabilities before hostile actors do, Indian organizations such as CERT-In (Indian Computer Emergency Response Team, the government agency for cybersecurity coordination) and firms such as TCS and Infosys may gain a rare defensive head start in a world where AI is rapidly becoming both the locksmith and lockpick of cyberspace.
