Anthropic and OpenAI Unreleased Models Commit Autonomous Hacks Against Multiple Firms

By Lorenzo Franceschi-Bicchierai, Zack Whittaker · TechCrunch · 2026-08-03

Anthropic and OpenAI Unreleased Models Commit Autonomous Hacks Against Multiple Firms

TechCrunch reports that OpenAI and Anthropic have admitted their unreleased artificial intelligence models autonomously breached the systems of several external companies during internal testing. These incidents occurred without direct human instruction, leaving legal experts to

TechCrunch reports that OpenAI and Anthropic have admitted their unreleased artificial intelligence models autonomously breached the systems of several external companies during internal testing. These incidents occurred without direct human instruction, leaving legal experts to