Google, OpenAI, Microsoft and 14 other technology companies announced the establishment of the Coalition for Safe Artificial Intelligence (CoSAI). The initiative aims to solve the “fragmentation problem of AI security” by providing open source methods, frameworks and tools.
Other founding members include Amazon, Intel, Nvidia, IBM, Cisco, Paypal, Anthropic, Cohere, Chainguard, WIZ, and GenLab.
CoSAI is defined as an open source community led by a Project Management Committee responsible for advancing and managing its overall technical agenda, and a Technical Steering Committee composed of AI experts from academia and industry to oversee its work processes.
The scope of work of CoSAI includes building, integrating, deploying, and operating AI systems securely, with a focus on mitigating risks such as model theft, data poisoning, hint injection, abuse at scale, and inference attacks. The project aims to develop comprehensive security measures that address typical and unique risks in AI systems.
CoSAI will initially form three workstreams:
- Software supply chain security for AI systems: Enhancing composition and provenance tracking to protect the security of AI applications.
- Preparing defenders for the changing cybersecurity landscape: Addressing investment and integration challenges in AI and traditional systems.
- AI safety governance: Develop best practices and risk assessment frameworks for AI safety.
There are plans to add more workflows over time.
