Clipped from: https://www.financialexpress.com/business/banking-finance/govt-bars-mythos-testing-in-banks/4222416/
Stance may be reviewed after US sandbox results on frontier AI model
India Bars Banks from Anthropic’s ‘Mythos’ AI: Pre-emptive Ban on Unreleased Model Over Cyber Threats
India has taken a pre-emptive stance on Anthropic’s unreleased frontier AI model Mythos, indicating that banks and insurance companies will not be allowed to test or deploy it until its safety is established, even as authorities step up preparedness against potential cyber risks, sources told Fe.
Defensive Pre-emption
The guidance does not amount to a ban on an existing deployment, as Mythos has not yet been made available outside a limited sandbox environment in the United States. Instead, it signals that any future access to the model will be subject to regulatory clearance.
“No stress testing with Mythos. We will not allow Mythos to be tested on our banks and insurance companies,” a senior official said. “How do we know whether it is safe or not?”
Officials said concerns stem from the model’s potential to significantly accelerate offensive cyber capabilities. Mythos is seen as capable of compressing months of cyber intrusion work into hours, raising the risk that automated attacks could outpace existing defence systems. The risk, they added, is less about systemic takeover and more about the speed and scale at which vulnerabilities could be exploited.
At present, Mythos is being used in four or five major US banks under what officials described as a sandboxing arrangement. Anthropic, along with US authorities, has not provided access to other countries, including India. The official said India would take a view on any future engagement after results from these sandbox tests are shared and access is offered.
“As of now, there is no need to worry, but we need to be very alert to the potential threat Mythos may pose to the country,” the official added.
Unprecedented Speed
The crisis from AI tools such as Mythos was reviewed at a meeting on April 23 chaired by Finance Minister Nirmala Sitharaman and Electronics and IT Minister Ashwini Vaishnaw, with participation from scheduled commercial banks and other stakeholders. Sitharaman said the emerging threat from advanced AI models is unprecedented and requires heightened vigilance, preparedness and coordination across financial institutions.
She also advised the Indian Banks’ Association to develop a coordinated institutional mechanism to respond to such risks. Banks have been directed to strengthen monitoring systems and hire specialised cybersecurity professionals and agencies.
Separately, the RBI is in discussions with global regulators, domestic lenders and the government to assess risks linked to frontier AI systems. Recent reports have described Mythos as a highly capable tool for identifying software vulnerabilities, analysing systems and potentially generating exploits, prompting regulators to prioritise defensive readiness even before any domestic exposure.
aniltikotekar4sme.com 