Indian Researchers Exploit OpenAI via Anthropic Model, Exposing AI Security Gaps

Indian researchers hacked OpenAI using Anthropic's model, highlighting AI security risks and vulnerabilities.

September 23, 2026 One Min Read

POLICY WIRE — New Delhi, India — A team of Indian cybersecurity researchers successfully exploited OpenAI’s systems using a rival AI model from Anthropic, raising concerns about the growing risks associated with advanced artificial intelligence technologies.

The three researchers, part of a company called Hacktron, used Anthropic’s Claude Opus 5 model to access user accounts on OpenAI’s community forum. The breach, which occurred in late July, demonstrated how rapidly AI capabilities evolve and how these advancements can be leveraged for both ethical and malicious purposes.

📄 POLICY WIRE WHITEPAPER PUBLISHED: PAKISTAN’S NATIONAL SECURITY POLICY PRIORITIES

👉 READ BY CLICKING HERE

Mohan Pedhapati, one of the researchers, described the process as a “force multiplier” that significantly reduces the time and effort required for complex cyber operations. He emphasized that the speed at which AI models improve creates new challenges for securing digital infrastructure, particularly when companies rely on third-party services that may have vulnerabilities.

Reporting by Policy-Wire (PW)

Author

Editorial Desk

Policy Wire is your go-to editorial hub for insightful perspectives and informed analysis on pressing policy issues, both regionally and globally. Over the years, Policy Wire has collaborated with leading international news agencies—including AFP, AP, Reuters, and others—to provide fact-based reporting and nuanced commentary. With a commitment to clarity and credibility, we aim to bridge the gap between complex policy debates and public understanding, ensuring that our readers stay informed on the issues shaping the world today.