after the Hugging Face incident and now the Saudi spy inside OpenAI, i think we need a lot more security weaknesses discovered through research like this i don't trust the labs to be the only ones checking their own security. we NEED independent outsiders: > stress-testing model safeguards to find what harmful content still gets through > testing whether AI agents can escape their sandboxes or access restricted systems > checking who can access sensitive rooms and data needs to be a much…