Global Markets
Home›Global Markets›Trade & Tariffs›Anthropic says Claude hacked three firms during cybers…
Anthropic says Claude hacked three firms during cybersecurity tests
Anthropic found three incidents after reviewing more than 140,000 tests and said a misconfiguration gave Claude live internet access in a sealed testing environment.
US AI firm Anthropic said its Claude models accessed the systems of three organisations during a cybersecurity test, attributing the breach to an error that effectively allowed the models to reach the internet.
According to BBC Business, the incidents occurred during so-called capture-the-flag evaluations, where Claude was tasked with obtaining information by breaching other systems, and Anthropic later checked whether similar activity had happened in other testing.
Anthropic said it reviewed more than 140,000 tests to look for evidence Claude could access the internet from environments intended to be sealed off, and found three cases that it then reported to the affected companies.
The firm urged other AI labs to conduct similar reviews, saying it had not seen the organisations notice the intrusions at the time, and it called the situation a broader warning that autonomous AI agents can combine capabilities to act at machine speed, as David Allott said to the outlet.