In brief
Anthropic has disclosed three incidents in which Claude compromised real world companies during cybersecurity evaluations.
A testing error gave the models internet access despite their being told they were operating in isolated environments.
The company says the incidents were caused by failures in testing infrastructure, not deliberate attempts by the AI to escape.
A week after OpenAI disclosed that its AI models escaped a locked testing environment and breached Hugging Face, and a day after admitting that its own AI models escaped containment, Anthropic revealed on Thursday that several versions of its Claude AI model also compromised three unnamed real-world companies after a misconfiguration gave the AI access to the open internet.Anthropic uncovered the incidents after reviewing more than 141,000 cybersecurity evaluation runs launched in response to the OpenAI disclosure.
