FM Sitharaman assesses Mythos risk on banking sector; India pings Anthropic for answers

A senior official said a report stating that Mythos, not yet widely released owing to its capabilities, was accessed by unauthorised individuals, had raised significant concerns.

“What is being broadly discussed is the basis on which the government and regulators, like the RBI, can build safety and security around such AI models. We are also talking to our own AI model developers and others because they understand how such systems are made and can be guarded against,” the official said.

Earlier this month, Anthropic unveiled a preview of Mythos, calling it the most powerful model it has ever built. So powerful, in fact, that unlike typical AI launches, the company is deliberately keeping the technology away from the public for now. The reason behind that is the model’s unprecedented ability to autonomously identify serious vulnerabilities in widely used software and infrastructure, posing significant cyber security threats.

Anthropic has said Mythos can outstrip all but the most skilled humans at identifying and exploiting software vulnerabilities, and in tests, the model found critical faults in every widely used operating system and web browser.

The model, which scored a staggering 73% on expert-level hacking tasks in independent testing by the United Kingdom’s AI Security Institute – tasks which no previous AI model could complete – has sent shockwaves across governments in the world, who are now scrambling to shore up defences against cyber attacks on critical sectors.

It is learnt that the Indian government has also reached out to Anthropic’s senior leadership in the US through the Ministry of External Affairs.

“We are talking to them to understand the issues surrounding Mythos, what their concerns are, and what governments can do to mitigate potential risks,” a senior official said.

Queries sent to the Finance Ministry, IT Ministry and Anthropic by The Indian Express did not elicit a response.

After Anthropic announced a staggered and controlled release of Mythos earlier this month, the IT Ministry set up the AI Governance and Economic Group (AIGEG) – a high-level inter-ministerial body – to serve as the apex mechanism for coordinating the country’s artificial intelligence policy and governance framework. A separate Technology and Policy Expert Committee (TPEC) was also created, led by the IT Secretary, to provide expertise to the inter-ministerial governance group.

“Internally, the government is also discussing next steps, and how it can increase protection for key sectors. We are also tracking a recent update that Mythos may have been accessed by certain unauthorised people, the situation is changing everyday. We are watching this very closely, and agencies like the Cert-In are also involved. We have sensitised chief information security officers of all companies that are operating in the critical information sector, including the banking sector and digital public infrastructure,” an official said.

The capability, the threat

Governments across the world are moving to step up defences against cyber attacks given that Mythos has the ability to autonomously identify serious vulnerabilities in widely used software and infrastructure, posing significant threats. It is also the reason why Anthropic is keeping the technology away from the public for now.