The Indian Computer Emergency Response Team (Cert-In) has operationalised an artificial intelligence (AI) war room to work with AI models that have 60-70 per cent capability of frontier AI models such as those of Anthropic’s Claude Mythos or OpenAI’s ChatGPT 5.6, among others, the secretary for the Ministry of Electronics and Information Technology, S Krishnan, said on Monday.
“They (Cert-In) have created sandboxes and testing environments where they are trying various elements with these AI models and identifying the possible vulnerabilities, so that we have, in a sense, done a dry run for whenever we have access to these models,” Krishan said during the launch of the second edition of Cert-In’s Digital Threat Report.
The idea, Krishnan said, was also to be prepared and to check for all vulnerabilities and potential patches as and when Indian companies and agencies gained unhindered access to these frontier AI models.
Apart from that, this war room will also ensure that a significant proportion of the development of indigenous AI models and protection mechanisms is done without relying heavily on these frontier general-purpose models, he said.
“There is a risk, of course. The government is concerned that we should have access. The processing of sensitive elements will have to be done on premises within the government,” Krishnan said.
In India, access to Anthropic’s latest general-purpose model Claude Mythos was initially extended to a few government agencies such as the Indian Cyber Crime Coordination Centre (I4C), Cert-In, the National Critical Information Infrastructure Protection Centre, and the Digital Intelligence Platform of the Department of Telecommunications as part of the expansion of Project Glasswing.
However, access was later halted after the US imposed restrictions on foreign persons’ access to Claude Mythos. The government is in talks with the US counterparts and the respective companies to restore India’s access to these models, Krishnan said.
