Anthropic's new AI model exposes fresh risks, flaws for cybersecurity, IT services

New Delhi: A powerful new AI model is forcing govts, banks, and technology firms to rethink the rules of cybersecurity - and in India, the stakes may be even higher.

Claude Mythos, developed by Anthropic, has demonstrated the ability to autonomously detect and exploit software vulnerabilities, including flaws that have persisted for decades. Early tests revealed that the model could identify long-standing weaknesses and simulate complex, multi-step cyberattacks, prompting the company to restrict its wider release. Anthropic CEO Dario Amodei highlighted the shift, noting that AI systems are now capable of finding vulnerabilities "that humans have missed", a signal of how quickly the cybersecurity landscape is changing.

US Treasury Secretary Scott Bessent reportedly convened a meeting with top bank executives - including leaders from JPMorgan Chase, Goldman Sachs, Citigroup, BoA, and Morgan Stanley - to assess the risks posed by such advanced AI systems.

That concern is not theoretical. According to Jaydeep Singh, GM for India at Kaspersky, the emergence of such systems represents a turning point not just for security professionals, but for everyday users. "We have been closely monitoring how AI is reshaping the threat landscape, and Claude Mythos represents a moment that every user, not just the cybersecurity industry, needs to understand," Singh said.

The dual-use nature of AI is at the heart of the concern. The same capability that strengthens defences can just as easily be weaponised. "The same capability that finds a 27-year-old vulnerability in hardened infrastructure is the capability that, in the wrong hands, turns every unpatched system into an open door," Singh added.

Cybersecurity firm Check Point Software Technologies echoed the warning. Sundar Balasubramanian, MD, India and South Asia, for Check Point, says, AI is "dramatically lowering the barrier to entry for cyber attackers," enabling even less-skilled actors to identify and exploit vulnerabilities. He added that defensive tools can be repurposed offensively, compressing the traditional gap between attackers and defenders.

Jayant Saran, partner, Deloitte India, described this as a "changed reality," where organisations must prepare for risks that were previously invisible. He called AI a "double-edged sword...that cannot be reversed," highlighting an accelerating race between those securing systems and those attempting to break them.

In India, the risks are amplified by scale. From UPI to banking and govt platforms, millions depend on digital infrastructure - much of it built on legacy systems. These systems are often slower to patch, harder to monitor, and lack continuous threat intelligence, creating what Saran called an "asymmetric risk exposure." Singh pointed out that this gap is especially critical in India, where legacy infrastructure serves hundreds of millions.

Beyond cybersecurity, ripple effects could reach financial markets. Analysts say models like Mythos could automate parts of software development, testing, and security - core functions of IT services industry.

While disruption may be gradual, labour-intensive outsourcing models could face pressure, while firms embracing AI may benefit.