Hugging Face's security team was unable to use mainstream AI tools for their investigation because safety filters blocked the analysis of exploit-related data