Home / Technology / India, Anthropic AI Deal Hits Snag Over Data Location
India, Anthropic AI Deal Hits Snag Over Data Location
27 Jul
Summary
- India wants AI model testing within its borders, not on foreign clouds.
- Anthropic's closed model, Mythos, complicates on-shore deployment.
- Disagreement delays India's access to advanced AI security initiative.
India's pursuit of sovereign AI infrastructure has led to a disagreement with AI firm Anthropic, potentially delaying access to its advanced Mythos model. India prefers that AI model testing, including that of Anthropic's frontier models, occur on infrastructure hosted within the country to ensure data remains under its jurisdiction.
Anthropic had initially proposed routing code through its cloud for testing, a model India found unacceptable due to data residency concerns. While the US eased export controls on Anthropic's models in June 2026, allowing broader access in the US, Mythos has not yet been rolled out internationally. This has impacted India's application for access under Project Glasswing, Anthropic's security initiative.
The core of the issue lies in Anthropic's proprietary model, Mythos, which is unlikely to be shared directly with clients. The alternative would involve Anthropic deploying the model on Indian infrastructure while maintaining operational control, a framework that is still being finalized. Consequently, India's participation hinges on resolving both regulatory clearances and deployment arrangements.
