Anthropic says its AI models hacked 3 organizations on their own during tests
Rival firm OpenAI last week disclosed similar incidents involving its models.