Why did OpenAI’s and Anthropic’s AI models hack other companies?

OpenAI and Anthropic say their models broke into other companies’ systems during testing, raising security concerns amid a heated debate over how to regulate AI.