Weeks after Washington ordered Anthropic to pull its most advanced AI models offline around the world, the US has assured India that access to the technology, once granted, will not be suddenly withdrawn — a commitment that Indian IT Minister S. Krishnan said it came from US officials during talks with India.
Krishnan, who was speaking to reporters after a bilateral meeting with US Under Secretary of State for Economic Affairs Jacob Helberg on Wednesday, said Washington sought to explain the thinking behind its decision to suspend Anthropic’s Mythos 5 and Fable 5 models globally, while separately committing that already extended access to trusted partners would continue.
“With regard to AI models, the US concern is mainly how these models will be used, how powerful they are, and what is the potential impact. They have been looking for a review mechanism for some of these models internally before they are released,” Krishnan said. But I think there was an understanding – and something they certainly mentioned – which is that access to technology, once provided, is not going to be cut off. That was a guarantee,” he said, adding that India would have to wait and see how things pan out in practice.
The exchange carries weight for India. Anthropic expanded its restricted Project Glasswing program to India in early June, giving Indian organizations access to the Mythos Preview model for cybersecurity testing – among the first countries to receive it. Days later, Washington export control directives forced Anthropic to disable both Mythos 5 and public-facing Fable 5 for every user around the world, including India, after the government said it learned of the model’s potential “jailbreak.”
Anthropic quickly complied with the US government order, saying it could not reliably separate foreign nationals from other users to comply with the order in any other way.
For his part, Helberg was publicly measured, saying only that the two governments would continue to talk. “These are very sensitive national security discussions that are not entirely appropriate for public consumption. But I think both sides really understand each other’s views, and we very much intend to continue to take a gradual, measured approach to how we launch Anthropic Models in a way that is safe, both for ourselves but also for our Indian counterparts as well as for all of our trusted partners in terms of our critical infrastructure, our energy grid, and so we will continue to make sure that we do the hard work in having those conversations,” he told reporters on the sidelines of the conference. summit.
The Krishnan-Helberg exchange was unveiled against the backdrop of the second Pax Silica Summit, a US-led initiative launched in December to align allies and partners in semiconductors, critical metals and AI supply chains – clearly conceived, in Washington parlance, as a counterweight to China’s dominance of supply chains themselves.
India signed the framework in February, on the sidelines of the AI Impact Summit in New Delhi, and Krishnan is representing the country at a meeting this week, with sessions on AI infrastructure, security and critical minerals.
Washington used the summit to expand the bloc further, announcing nine new signatories on Wednesday: Argentina, Germany, the Netherlands, Chile, Costa Rica, Panama, Greece, Kazakhstan and the European Union. India, according to the Ministry of External Affairs, is among the 17 countries listed as signatories to the Pax Celica Declaration.
At the summit’s opening ceremony, US Deputy Secretary of State Christopher Landau said future technologies were too important to be “left vulnerable to coercive policies and markets” — a reference, without naming China directly, to Beijing’s role in critical metals and semiconductor supply chains.
Landau said he was confident that no “state-directed competitor” would be able to keep pace with the private sectors of the United States and its partners. “Our combined capabilities are something no command economy can match on its own,” he said.
