OpenAI and Anthropic AI Models Breached Other Systems During Testing, Raising Security Concerns | HyperFast