OpenAI ‘ethically hacked’ with help of Anthropic’s Claude chatbot

US cybersecurity researchers who conducted hack say ‘scope of what we could theoretically access was huge’ Cybersecurity researchers have hacked into OpenAI with the help of Anthropic’s Claude chatbot, in the latest example of security issues at the company. A team at a US-based startup compromised a number of OpenAI employees’ ChatGPT accounts, starting a process that enabled them to access their target’s software cache – and potentially more. Continue reading...

Hacking into OpenAI poses significant security risks.

The ethical hack reveals large vulnerabilities in OpenAI's security, risking unauthorized access to sensitive AI systems and user data. While the hack was controlled, the potential for malicious exploitation suggests material concerns about societal safety and trust in AI services. Evidence is from a specific reported incident, supporting moderate confidence.

This describes a disclosed, ethical penetration test revealing a real vulnerability at OpenAI, not an active malicious breach. It signals ongoing security fragility at a major AI lab, which has mildly concerning implications, but the responsible disclosure framing and lack of harm limit severity.

The hack demonstrates a significant security vulnerability in AI systems. The potential for unauthorized access to sensitive data is high. This could erode trust in AI technologies.

SOURCE     The Guardian: Artificial Intelligence (tier 1)
   ↓
DOCUMENT   7d59d59c-7b65-44f0-b83a-3d93e5546caf
           https://theguardian.com/technology/2026/sep/18/openai-hacked-anthropic-claude-chatbot
   ↓
EVIDENCE   2 extracted excerpts
   ↓
MODEL RUN  4 runs, methodology 1.0.1
   ↓
SCORE      -52  (Adverse)
   ↓
CONFIDENCE 70%