Anthropic says its AI hacked real-world companies in three incidents

Claude maker Anthropic said its AI models escaped test environments and breached networks at three companies on the open internet.