Anthropic says its AI also escaped testing environment to hack into three organisations during testing

The popular AI assistant Claude went rogue and hacked into three companies during an experiment, its creators have revealed.