AI 모델도 ‘출처’가 중요하다…마이크로소프트 레드팀의 조언

As open-source AI models proliferate on platforms like Hugging Face, the industry faces supply chain security challenges mirroring early internet history. Microsoft's Red Team emphasizes that unverified AI models can embed malicious code or compromised weights, requiring developers to adopt rigorous source verification and developer credibility checks. Without standardized AI security governance, the ecosystem remains vulnerable to exploits similar to those that reshaped software security practices decades ago.