Anthropic reveals Claude "gained unauthorized access" to "real-world systems"

Anthropic's artificial intelligence model Claude "gained unauthorized access" to three outside organizations on three separate occasions during testing.

5 min readNew DelhiJul 31, 2026 08:47 AM IST Anthropic has paused its cybersecurity evaluations after discovering three incidents involving Claude AI models.(Image: Reuters)Amid the intensifying debate around autonomous AI systems and cybersecurity, frontier AI lab Anthropic, on Thursday, July 30, said that it had foun

The incident comes just days after rival OpenAI first revealed its models went rogue during security testing.

In a review triggered by OpenAI's Hugging Face incident, Anthropic discovered three of its AI models had breached real organizations during third-party evaluations.

The artificial intelligence firm Anthropic revealed Thursday its Claude model escaped an isolated testing environment at least three times and accessed the systems of three different organizations without a prompt to do so. Anthropic said in a blog post Thursday evening it reviewed more than 141,000 evaluations of Clau

The disclosure followed OpenAI’s report last week that its own artificial intelligence had hacked into the network of an online library.

Claude thought it was in a simulation. It was on the open internet, and the companies it broke into were real.

Claude thought it was in a simulation. It was on the open internet, and the companies it broke into were real.

Anthropic said it discovered three instances where its Claude AI models accessed the internet during an evaluation and accessed outside systems.

Some of Anthropic's most powerful models — including Mythos 5 and an internal research model — gained unauthorized access to real-world systems during pre-deployment cybersecurity testing, the company said Thursday. Why it matters: OpenAI's and Anthropic's latest disclosures show frontier AI models reaching real-world