US tech company Anthropic said three of its artificial intelligence (AI) models hacked three organisations during tests, days after rival OpenAI said rogue AI agents had attacked the networks of other firms. The company said its Claude AI model gained unauthorised access to systems during a cybersecurity exercise by connecting to the internet from isolated test environments.
Anthropic said it discovered the incidents after reviewing more than 140,000 tests following OpenAI's disclosure on 21 July that its agents had hacked another AI company, Hugging Face. The San Francisco-based firm said it has alerted the three companies that were hacked about the incidents.
Anthropic urged other AI labs to carry out similar reviews to better understand the risks posed by their models' capabilities.
Anthropic said in a statement, external that it is "approaching the fixes as if the responsibility were ours alone."
