The Coalition for Secure AI (CoSAI), proposed by the global open standards organization OASIS Open, has been officially established. Its goal is to provide industry practitioners and developers with the necessary guidelines and tools to help create secure-by-design AI systems. Members include tech giants such as Intel, Google, IBM, Nvidia, and Microsoft, among other organizations.
Intel, a founding member of CoSAI, stated that the coalition represents a significant industry collaboration, bringing together experts from industry, academia, and other related sectors. These experts will jointly develop and share comprehensive solutions, best implementation practices, methods, and tools for secure AI development and deployment.
In its initial phase, CoSAI members will focus on three key workflows. Firstly, they will address the security of the AI system software supply chain, enhancing supply chain composition and source tracking to ensure the security of AI applications.
Secondly, they will assist defense systems in preparing for the rapidly changing field of cybersecurity, addressing the investment and integration challenges of AI and traditional systems. Third, they will work on AI security governance to establish best implementation practices and risk assessment frameworks for AI security.
CoSAI complements the newly released Open Platform for Enterprise AI (OPEA) sandbox project, of which Intel is also a founding member. Launched by the Linux Foundation's LF AI & Data Foundation, OPEA aims to promote interoperability among diverse and heterogeneous ecosystems. It will help enterprises accelerate the deployment of secure and cost-effective generative AI, starting with Retrieval-Augmented Generation (RAG).
As AI rapidly changes the world, developers and adopters face the challenge of ensuring AI technology's security while adhering to inconsistent or isolated guidelines and standards. To overcome these challenges, developers must prioritize security during development and share practical methods. CoSAI's establishment is a significant step towards addressing these concerns and promoting a more secure AI ecosystem.
