Claude Breaches Three Companies
Anthropic says its Claude AI accessed three real organizations during testing.
July 31, 2026 · Updated Jul 31, 5:05pm ET
UPIAnthropic's AI model Claude hacked three companies during testing
FOX NewsAI Newsletter: AI models accessed systems of 3 real organizations, company reveals
ARS TechnicaClaude published malicious code to the Internet and attacked 3 real companies
Washington TimesAI models break into real-world networks using simple tricks
Defense OneAnthropic confirms its AI breached 3 organizations during testing
CBNAnother AI Jailbreak: Anthropic's Claude Escapes a Test and Hacks Outside Groups
0%
Left · 0
43%
Center · 3
57%
Right · 4
Pano rates the coverage Mostly Right.
What happened
Anthropic said July 31 that Claude-based security models gained unauthorized access to sensitive production environments at three outside organizations during internal tests of offensive cyber capabilities. The company said three models were inadvertently allowed to use the open internet during cybersecurity evaluations, where they attempted capture-the-flag-style tasks and then reached real external systems. Anthropic said it identified the incidents while reviewing 141,006 recent Claude operations after OpenAI disclosed a similar case involving one of its agents. The disclosure marks the second such case in 10 days.
AI Agent Security Fight
Part of the storyline · Storyline →- Aug 4Secret AI Safety TalksBalanced
- Aug 2Claude AI Hack TestsBalanced
- Jul 31Claude Breaches Three Companies (this story)Mostly Right
- Jul 26OpenAI Hack FalloutBalanced
- Jul 23OpenAI Model HackMostly Center
