Indian firms step up cyber checks over Anthropic AI

Indian enterprises are assessing cybersecurity risks after Anthropic’s AI model Mythos showed it could find software flaws in hours.

HDFC Bank said it is working with the Data Security Council of India to assess the impact and is in touch with Anthropic, while Zerodha said it already uses large language models for internal security reviews.

Experts said banks and telecom firms that rely on older systems could be more exposed, and India’s cyber rules were built for slower threats.

Financial regulators in the UK and Canada have begun urgent discussions, while faster automation could pressure parts of India’s IT services and cybersecurity work.

🔗 Source: The Economic Times

🧠 Food for thought

Implications, context, and why it matters.

Mythos brings a risky new kind of offensive cyber capability

Anthropic’s Mythos goes beyond a vulnerability scanner. It found thousands of critical flaws in leading operating systems and web browsers, then produced working exploits quickly 1.
The report says non-experts can use it well, which may cut the know-how needed for advanced cyberattacks 1.
Access stays limited to a small set of partners, including Google and Microsoft. The report ties the limits to fears that a public release could cause harm 1.

AI-driven attacks are pushing security teams to rethink defense

Tools like Mythos are speeding up a broader move toward AI-based defense. Palo Alto Networks has rolled out AI-driven security products aimed at blocking cyberattacks 2.
Security teams are also focusing on identity protection for human users, machine accounts, and AI agents. The report describes these identities as a growing entry point as companies expand cloud computing, automation, and AI use 3.
In India, the pressure also links to the Digital Personal Data Protection Act (DPDPA). Banks are likely to fall under Significant Data Fiduciaries (SDFs), which brings tighter compliance duties under the law, according to a Protiviti report 4.
Protiviti also cites survey results that 52% of organizations had a privacy breach in the past five years, while 42% had a fully defined privacy program 4.