Mythos, an advanced AI model developed by Anthropic, can compress weeks of hacking effort into hours. This is making banks nervous globally.
A new kind of threat is forcing governments around the world to act quickly. When Union Finance Minister Nirmala Sitharaman called top bankers into a high-level security meeting last month, the message was unusually stark. The threat on the table wasn’t a malware strain or a ransomware gang. It was an AI model. Called Claude Mythos, the system is widely seen as a potential inflection point in cybersecurity. Not because it attacks systems directly, but because it can find ways to break them faster than humans ever could.
The finance ministry described the risk as “unprecedented”. The choice of word caught attention. Governments rarely escalate language unless the risk cuts across the entire financial system.
What Is Mythos AI?
Mythos is an advanced AI model developed by Anthropic. Unlike conventional cybersecurity tools that detect known threats, Mythos is believed to identify unknown vulnerabilities-the so-called “zero-day” flaws.
These are the most dangerous bugs in software. No one knows they exist until they are exploited.
In theory, Mythos can:
- Scan systems and detect hidden weaknesses
- Suggest ways to exploit those weaknesses
- Compress weeks of hacking effort into hours
In fact, Anthropic itself has refused to release the AI model publicly. This alone signals how sensitive it is. However, the company has provided controlled access to select firms like Apple and Goldman Sachs. The concern around the AI model escalated further after reports suggested unauthorised access to the model may have already occurred.
Mythos AI: India Takes Pre-Emptive Action
India’s response to the threat is not reactive. It is pre-emptive. The government is already working on a coordinated framework that allows:
- Real-time threat intelligence sharing across banks
- Coordination with agencies like Indian CERT (Computer Emergency Response Team)
- Faster incident response across the financial system
Sitharaman has also pushed the Indian Banks’ Association to create a unified institutional mechanism. The idea is simple: if an attack happens, no bank should be fighting it alone. The Reserve Bank of India (RBI) is also holding consultations to get on top of the situation.
This mirrors global moves. The US government has, as per reports, engaged Wall Street banks on similar risks. Interestingly, there are also discussions about using such AI systems defensively within government networks.
How Mythos Could Impact Banks & Financial Systems?
The risks go far beyond a typical cyberattack. Sudiptaa Paul Choudhury of QNu Labs calls it a “watershed moment”, where cyber risk could evolve into a systemic financial threat. In extreme scenarios, vulnerabilities exposed by AI could disrupt payments, freeze ATMs, or even trigger liquidity stress and public panic.
Yogesh Jadhav, Group CTO at Choice International, frames the urgency bluntly: attack timelines are shrinking. What once took weeks-finding and exploiting a flaw-could now happen in a day. This fundamentally changes how banks tackle cyber threats — detect, patch, recover.
Mythos AI: From Prevention To Resilience
Experts are clear on one point. Prevention alone will not work anymore. Amit Nigam of FindiBANKIT argues that cybersecurity is no longer just a tech problem. It is a systems design and governance challenge. Banks need continuous monitoring, auditability, and human oversight built into their architecture.
Ayush Sarvaiya of Plus91Labs flags another issue: over-reliance on systems that are not fully understood. In banking, opaque AI decisions can create accountability gaps that are hard to fix after the fact.
The consensus is shifting toward:
- “Secure-by-design” systems
- Continuous threat detection
- Strong access controls and segmentation
- Human-in-the-loop decision making
- AI is both the threat-and the defence
There is an irony here. The same capabilities that make Mythos dangerous can also make financial systems stronger.
Piyush Bagaria of SalarySe notes that AI is already reshaping how institutions think about risk-moving from static rules to real-time, behaviour-driven systems. This has implications beyond cybersecurity, including fraud detection and credit decisioning.
Amit Chandel of Olyv adds that the real differentiator will not be the technology itself, but how responsibly it is deployed. Institutions that build with resilience, trust, and accountability will come out ahead.
