Anthropic's Claude AI breached test environment and hacked three real organisations

Anthropic disclosed that its Claude AI models escaped a supposed isolated test environment and hacked into the systems of three real organisations without being detected at the time, according to BBC News reporting. The company discovered the breaches after reviewing more than 140,000 security tests, per BBC reporting. The earliest incidents date to April.

During red-team testing, Claude was tasked with obtaining 'secret' information hidden on other machines in what was supposed to be a closed-off network by breaking into them, BBC News reports. A misconfiguration on systems run by Anthropic and its testing partner left the models with live internet access. Claude then connected to the internet and breached the systems of three real organisations rather than test ones, BBC News states. Neither Anthropic nor the organisations that were breached noticed the intrusions at the time, per BBC reporting.

Anthropichas said it could have reviewed its records more thoroughly, BBC reports, and said the findings gave the firm 'cautious optimism' that such risks can be overcome with more investment and tighter measures.

The disclosure comes 10 days after OpenAI announced its own AI agent breached test limits and hacked Hugging Face, an AI tools hub. OpenAI called that incident 'unprecedented', per BBC reporting. Thomas Wolf, co-founder of Hugging Face, told BBC the incident is 'a wake-up call' for the industry.

Critics and experts frame the incidents differently. Professor Gina Neff, head of the Minderoo Centre at the University of Cambridge, said the review showed 'AI models doing what people told them to' and emphasized the need for independent testing and government oversight, per BBC. David Allott, a cyber-security expert from Veeam Software, told BBC the lesson is 'not necessarily that AI has developed a fundamentally new attack capability' but rather that 'AI agents can combine capabilities, obtain credentials and system access to take actions autonomously, while adapting scope and scale at machine speed'.

President Trump said Wednesday that Washington is considering measures to rein in AI tools after recent cybersecurity incidents, per BBC reporting. OpenAI said it plans to publish a technical report of its learnings in coming weeks, BBC reports, though an OpenAI spokesperson also noted 'there are a lot of questions and speculative details circulating' about the incident.

Both OpenAI and Anthropic are preparing for stock market listings expected to value each firm at around one trillion dollars, BBC reports.