Introduction

  • TL;DR: Sentient AGI’s OML 1.0 enables verifiable model ownership by embedding 24,576 fingerprints in LLMs without degrading performance. The system, presented at NeurIPS 2025, establishes a cryptographic foundation for sustainable open-source AI monetization and ethical distribution.
  • OML (Open, Monetizable, Loyal) is a framework designed to reconcile open access with creator control in AI model distribution. Sentient AGI’s implementation focuses on embedding cryptographic fingerprints within models to enable provable ownership while maintaining model utility and openness.

OML 1.0’s Core Design

Fingerprint Embedding

OML 1.0 fine-tunes LLMs using secret query-response pairs, creating cryptographic “fingerprints.” These behave as immutable model signatures, enabling proof of genuine ownership against unauthorized replication. Each fingerprint acts like a model’s cryptographic DNA that can be verified without compromising the model’s functionality.

Why it matters: Fingerprint-based attribution makes open-source AI commercially viable while ensuring ethical transparency and enabling creators to maintain control over their models even after distribution.

AI-Native Cryptography and Optimistic Security

OML 1.0 is built on AI-native cryptography, leveraging the differentiability in neural networks to achieve verifiable yet non-removable security. The “Optimistic Security” model assumes honest use by default, but any violation triggers detectable fingerprint checks through specific query patterns.

Why it matters: This approach balances openness and loyalty, a key requirement for decentralized AI networks, allowing models to be freely distributed while maintaining verifiable ownership and usage tracking.

Ecosystem and Academic Validation

The OML framework was presented at NeurIPS 2025 with participation from researchers at leading institutions. The Sentient Protocol extends OML 1.0 with on-chain verification and profit-sharing mechanisms, creating an ecosystem where model creators can monetize their work while maintaining open access.

Why it matters: It sets the stage for transparent, fair AI ecosystems where creators retain verifiable rights and can build sustainable business models around open-source AI development.

Conclusion

OML 1.0 represents a significant advancement in AI model ownership and distribution. By embedding cryptographic fingerprints that are both verifiable and difficult to remove, Sentient AGI has created a practical framework for maintaining creator rights in open-source AI models.

Summary

  • OML 1.0 embeds cryptographic fingerprints at scale (24,576) with minimal performance impact
  • Enables verifiable model ownership and supports sustainable monetization of open-source AI
  • Demonstrates practical AI-native cryptography for trustworthy open collaboration
  • Framework presented at NeurIPS 2025 as a foundational technology for the evolving AI economy

#SentientAGI #OML #AIOwnership #ModelFingerprinting #NeurIPS2025 #OpenSourceAI #AICryptography

References

  1. “OML: A Primitive for Reconciling Open Access with Owner Control in AI Model Distribution” | arXiv | 2024-11 | https://arxiv.org/abs/2411.03887
  2. “OML and the Sentient Protocol” | Open AGI | 2024-11-02 | https://openagi.discourse.group/t/oml-and-the-sentient-protocol/1507
  3. “Sentient AGI OML 1.0 Fingerprinting” | GitHub | 2024-11-13 | https://github.com/sentient-agi/OML-1.0-Fingerprinting
  4. “The Sentient Foundation” | Sentient.foundation | 2024 | https://sentient.foundation