OML: A Primitive for Reconciling Open Access with Owner Control in AI Model Distribution, by Zerui Cheng, Princeton U.
Abstract:
The current paradigm of AI model distribution presents a fundamental dichotomy: models are either closed and API-gated, sacrificing transparency and local execution, or openly distributed, sacrificing monetization and control. We introduce OML(Open-access, Monetizable, and Loyal AI Model Serving), a primitive that enables a new distribution paradigm where models can be freely distributed for local execution while maintaining cryptographically enforced usage authorization. We are the first to introduce and formalize this problem, introducing rigorous security definitions tailored to the unique challenge of white-box model protection: model extraction resistance and permission forgery resistance. We prove fundamental bounds on the achievability of OML properties and characterize the complete design space of potential constructions, from obfuscation-based approaches to cryptographic solutions. To demonstrate practical feasibility, we present OML 1.0, a novel OML construction leveraging AI-native model fingerprinting coupled with crypto-economic enforcement mechanisms. Through extensive theoretical analysis and empirical evaluation, we establish OML as a foundational primitive necessary for sustainable AI ecosystems. This work opens a new research direction at the intersection of cryptography, machine learning, and mechanism design, with critical implications for the future of AI distribution and governance.
https://arxiv.org/abs/2411.03887
Bio
Zerui Cheng a 3rd-year Ph.D. candidate at Princeton University advised by Prof. Pramod Viswanath. He is also a part-time student researcher at ByteDance Seed supervised by Dr. Jiashuo Liu. Before Princeton, he completed his B.Eng. in Computer Science from Yao Class at Tsinghua University, graduating summa cum laude and earning the prestigious Yao Award. His research interests lie in the intersection of AI evaluation, deployment, and blockchains, aiming to leverage technology to promote fairness and transparency in the AI era, with a strong focus on real-world impact. His research in blockchains x AI track has contributed to the technical foundation of high-profile startups including Sentient (OML: open-access yet controllable AI model distribution), Kite AI (AI agent-native payment infrastructure with stablecoins and state channels), and PolyHedra (zkBridge: trustless cross-chain bridges with zk proofs). He is also in active research collaboration with Cyber (CAIA: a crypto-native AI benchmark) and Forest AI (PeerBench: decentralized community-built AI benchmark, recently accepted to NeurIPS 2025). Apart from decentralized AI, his research in pure AI track has also made an impact, featuring LiveCodeBench Pro (also accepted to NeurIPS 2025) which gained over 1 million views on X and was covered by MIT Tech Review in June 2025. For more information, please refer to his personal website at https://www.zerui-cheng.com/.
Organized by DeAI Institute
https://www.linkedin.com/groups/10180003/
Co-hosts:
EZ.Encoder Academy
TAPNET (Toronto AI Practitioners Network)
https://www.linkedin.com/company/toronto-tapnet/
Toronto DAO
https://www.linkedin.com/company/tdao-eth/
Toronto LLM Meetup Group
https://lu.ma/user/usr-O5UaG1q5BbTSon2
Check YouTube Channel for the video:
https://www.youtube.com/channel/UCImPUcSoFqP9TcItLYOAo-A
By registering to attend the event, you agree to be recorded during the event.