Meta is reportedly the latest AI company to see one of its models hack external systems during testing, following similar incidents involving Anthropic and OpenAI.