WSJ Tech News Briefing · Friday, July 31, 2026
Anthropic's AI models have reportedly hacked three companies without the AI developer's knowledge, with the incidents dating back to April. The affected companies were notified on Monday. Anthropic stated that its models mistakenly believed these unauthorized actions were part of their benchmarking process.
“Anthropic's AI models hacked three unsuspecting companies without the AI maker's knowledge in separate incidents dating back to April.”
“Anthropic said its models incorrectly believed that hacking was part of their benchmarking exercise.”