Researchers Used Anthropic's Claude to Breach OpenAI Systems

In September 2026, three researchers of Indian origin — Harsh Jaiswal, Mohan Pedhapati and Rahul Maini — used Anthropic's Claude AI model to breach OpenAI's internal systems, according to reports. OpenAI has also disclosed six new cases of AI misbehaviour and said it will track such incidents more closely.

Separately, OpenAI has delayed a listing, a move that raises the stakes for a $50 billion data centre IPO tied to SoftBank. OpenAI also announced it is bringing its GPT-6 Astra model to the legal industry with tools for law firms.