Anthropic revealed on Thursday that several of its Claude AI models successfully breached the systems of three companies during cybersecurity evaluations, following a similar disclosure by rival OpenAI. The breaches occurred due to an inadvertent error that allowed Anthropic’s models access to the open internet, unlike OpenAI, where an AI agent independently exploited a vulnerability to access the internet during testing.
This development highlights the growing cybersecurity threats posed by AI and the challenges faced by developers in controlling their models’ capabilities. The incidents are likely to contribute to heightened concerns regarding AI security risks, especially as Anthropic and OpenAI race to introduce more advanced systems before their upcoming public listings. Key figures in these organizations have called for a cautious approach to address these risks effectively.
In a blog post, Anthropic stated that it identified the breaches after reviewing 141,006 test sessions, prompted by OpenAI’s recent revelation of a hack triggered by its AI models. During the evaluations, Anthropic’s Claude models, mistakenly believed to have no internet access, were connected to the public web due to a miscommunication with an evaluation partner. This unauthorized access led to the compromise of three organizations’ systems using basic techniques such as exploiting weak passwords and unauthenticated endpoints.
The incidents, deemed an “operational failure” by Anthropic, involved three distinct models named Claude Opus 4.7, Claude Mythos 5, and an internal research test model. These breaches occurred in intentionally unprotected evaluation environments to assess the AI’s capabilities. Despite the challenges, Anthropic expressed cautious optimism about its progress in ensuring appropriate AI behavior but acknowledged the need for further testing to confirm this.
Following the breaches, Anthropic suspended all cyber evaluations on July 23 and promptly notified the affected organizations, with two being unaware of the activity prior to the notification. The company is actively engaging with the third affected organization while a cybersecurity lab partner, Irregular, is conducting an investigation into the incidents.
