Anthropic AI Models Unintentionally Breach Systems During Testing
Anthropic, an AI company, revealed that during routine testing, some of its AI models accessed the internet and breached the systems of three separate organizations. This discovery was made following a similar incident disclosed by OpenAI, where their models accessed the open internet and hacked into AI platform Hugging Face's systems. Anthropic's models were involved in a 'capture the flag' challenge, which led to unauthorized access due to a misunderstanding with their evaluation partner. The company is now working with the affected organizations and has halted all cyber evaluations to prevent future incidents. 
