July 19, 2024 – At the Aspen Security Forum held yesterday, 14 leading companies in the AI field, including Google, Microsoft, OpenAI, Amazon, NVIDIA, and Intel, announced the formation of the Coalition for Secure AI (CoSAI). This alliance is aimed at enhancing the security of artificial intelligence systems.
Hosted by the OASIS global standards body, CoSAI is an open source initiative designed to provide AI practitioners and developers with the guidance and tools necessary to create securely designed AI systems. The coalition intends to establish a collaborative ecosystem that shares open source methodologies, standardized frameworks, and tools.
The founding sponsors of CoSAI include Google, IBM, Intel, Microsoft, NVIDIA, and PayPal, among others such as Amazon, Anthropic, Cisco, Chainguard, Cohere, GenLab, OpenAI, and Wiz.
Initially, CoSAI will focus on achieving three primary objectives: developing best practices for AI security, addressing challenges in the field of AI, and ensuring the safety of AI applications.
The scope of CoSAI’s work encompasses the secure construction, integration, deployment, and operation of AI systems. The alliance will prioritize mitigating risks such as model theft, data poisoning, prompt injection, scale abuse, and inference attacks.
CoSAI operates as an open source community led by a project council and a technical steering committee. The project council is responsible for advancing and managing the overall technical agenda, while the technical steering committee, consisting of AI experts from academia and industry, oversees the workflow.
According to a statement by Heather Adkins, Vice President of Security at Google, “AI has been developing for many years, presenting continuous potential for defenders while also recognizing the opportunities it brings for adversaries.” She further added, “CoSAI will assist organizations of all sizes to integrate AI safely and responsibly, helping them harness the advantages of AI while minimizing risks.”
