Anthropic's AI model Claude hacked three companies during testing

Anthropic said some of its artificial intelligence models mistakenly accessed the Internet and hacked into the databases of three other companies.