Formation of the 'Open Secure AI Alliance' for AI Safety and Security

Open source software is a core pillar of the global economy across cloud, finance, manufacturing, and more, and its importance is especially high in the field of cybersecurity. The Open Secure AI Alliance builds on the achievements of the Linux Foundation's Akrites initiative and OpenSSF, aiming to use open technology to address and disclose vulnerabilities in the AI era.

Open models and open harnesses are essential to AI security. They democratize defense capabilities, increase transparency, and enable cyber defense while protecting data. Closed models alone are limited in their ability to protect infrastructure, and open frontier tools that defenders can directly inspect and adapt are needed.

A recent security incident at Hugging Face demonstrated this necessity. When a closed AI tool failed to distinguish between attackers and defenders and blocked forensic analysis, Hugging Face was able to run the open-weight model GLM 5.2 on its own infrastructure to analyze more than 17,000 actions and block the intrusion.

Major companies participating in this alliance include:

• NVIDIA, Microsoft, Adobe, Cisco, IBM, Red Hat
• NAVER, SK Telecom, Samsung (Korean companies included in the list of participants)
• CrowdStrike, Palo Alto Networks, Cloudflare, and others

While there is a risk that open models could be misused, this is a problem that also occurs in closed systems. Therefore, the right response is not to reject open models, but to combine strong protective measures, clear rules, rigorous evaluation, and rapid recovery systems so that defenders have tools they can trust.