India sounds alarm, demands fair access to Anthropic's Mythos AI

India has opened talks with the US and Anthropic after being excluded from early access to Mythos AI, a model seen as a major cybersecurity threat. The Centre is now pushing for fair access while strengthening protection of banks, telecom and power networks.

In Short

• India seeks fair access to Anthropic’s Mythos AI
• Govt flags risks to banking, telecom, power grids
• No Indian firm included in early Glasswing access

India has stepped up talks with the United States and Anthropic after being left out of early access to the company's powerful new AI model, Mythos, which has triggered fresh concerns around cybersecurity and critical infrastructure safety. Officials familiar with the matter told Economic Times that the Centre is now exploring ways to ensure Indian companies get fair access to the model while also protecting key systems such as banking networks, telecom infrastructure and power grids. The government is said to be working on both "mechanisms" and "logistics" as part of ongoing bilateral discussions with the US administration.

What is Mythos AI and why India is worried about it

The urgency comes from what Mythos is believed to be capable of. Anthropic has described the model as advanced enough to rapidly detect and exploit zero-day vulnerabilities, which are basically the hidden software flaws that attackers can misuse before developers even know they exist. Such weaknesses can be found in operating systems, browsers and other widely used digital tools.

Finance Minister Nirmala Sitharaman just confirmed that the matter is being handled at the highest levels. Speaking at the ET Awards for Corporate Excellence, she said the Ministry of Electronics and IT was actively engaged with the US administration, Anthropic and vendors already testing the model. "The cyber challenge we have because of the Mythos, is going to be a big one," she said.

Her remarks came shortly after a high-level meeting involving Sitharaman, IT Minister Ashwini Vaishnaw, banks and key cyber agencies. Officials say India wants to avoid a situation where access is limited to select private players while others are excluded, especially if the technology has implications for national security.

At the moment, Anthropic has not opened Mythos for public use. Instead, access has reportedly been granted to around 40 companies under a programme called Project Glasswing. Most of these firms are based in the US, and no Indian company was included in the first list.

That has raised concerns in India’s policy and technology circles. Kazim Rizvi, founding director of think tank The Dialogue, said India should urgently seek participation in Project Glasswing or an equivalent arrangement. But, India’s concern is not just about access today, but what comes next. Officials believe more AI companies could soon launch similar models without prior warning, creating new cybersecurity risks across countries.

For that reason, the Centre has reportedly asked CERT-In, the National Critical Information Infrastructure Protection Centre (NCIIPC), and financial sector stakeholders to accelerate preparedness measures. These include reviewing vulnerabilities in sensitive systems and strengthening cyber defence frameworks.

Experts say AI tools like Mythos could be used in two ways — by defenders to identify flaws quickly, or by malicious actors to launch more advanced attacks. That dual-use nature is what makes the technology especially sensitive.

There are also reports that unauthorised users may have briefly gained access to the model in private online forums, adding to fears that containment may not be easy.

- Ends