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

This collaborative approach highlights a growing trend where leading labs leverage competing AI models to pressure-test their own systems for security vulnerabilities. By utilizing Claude to stress-test GPT-4, OpenAI demonstrates that transparency and cross-platform cooperation are becoming essential tools for robust AI safety.