CoSAI Launched: Founding Members Unite for Secure AI

AI's rapid growth demands a robust security framework and standards that can keep up. That's why we introduced the Secure AI Framework (SAIF) last year, knowing it was just the beginning. To make any industry framework work, you need teamwork and a place to collaborate. That's where we're at today.

At the Aspen Security Forum, we're excited to announce the launch of the Coalition for Secure AI (CoSAI) with our industry buddies. We've been working hard for the past year to bring this coalition together, aiming to tackle the unique security challenges AI brings, both now and in the future.

CoSAI's founding members include big names like Amazon, Anthropic, Chainguard, Cisco, Cohere, GenLab, IBM, Intel, Microsoft, NVIDIA, OpenAI, PayPal, and Wiz. We're setting up shop under OASIS Open, the international standards and open-source group.

Introducing CoSAI’s Inaugural Workstreams

As everyone from individual developers to big companies works on adopting common security standards and best practices, CoSAI will back this collective effort in AI security. Today, we're kicking off with the first three areas the coalition will focus on, working hand-in-hand with industry and academia:

1. Software Supply Chain Security for AI systems: Google's been pushing to extend SLSA Provenance to AI models, helping to figure out if AI software is secure by tracing its creation and handling through the supply chain. This workstream aims to boost AI security by offering guidance on checking provenance, managing risks from third-party models, and assessing the full AI application provenance, building on the existing SSDF and SLSA security principles for both AI and traditional software.
2. Preparing defenders for a changing cybersecurity landscape: Dealing with daily AI governance can be a real headache for security folks. This workstream will create a defender's framework to help them spot where to invest and what mitigation techniques to use to tackle the security impacts of AI. The framework will grow with the rise of offensive cybersecurity advancements in AI models.
3. AI security governance: Governing AI security needs new resources and an understanding of what makes AI security unique. CoSAI will develop a list of risks and controls, a checklist, and a scorecard to help practitioners assess their readiness, manage, monitor, and report on the security of their AI products.

Plus, CoSAI will team up with groups like Frontier Model Forum, Partnership on AI, Open Source Security Foundation, and ML Commons to push forward responsible AI.

What’s Next

As AI keeps moving forward, we're all in on making sure our risk management strategies keep pace. We've seen great support from the industry over the past year to make AI safe and secure. Even better, we're seeing real action from developers, experts, and companies of all sizes to help organizations use AI safely.

AI developers need—and end users deserve—a security framework that's up to the challenge and responsibly seizes the opportunities ahead. CoSAI is the next big step in this journey, and we'll have more updates coming soon. To find out how you can support CoSAI, check out coalitionforsecureai.org. In the meantime, head over to our Secure AI Framework page to learn more about Google's work on AI security.