Skip to content Skip to footer

Smart Privacy for Web3 & AI

Learn how the Oasis Foundation looks to help build the decentralized AI industry

We are in the midst of a revolution with generative AI (GenAI) and Large Language Models (LLMs) touching and disrupting every walk of life at a breathtaking pace. The need of the hour is for technology that can realize decentralized, trustless and democratic development and deployment of AI systems.

Ever since the founding of Oasis in 2018, with academic roots firmly embedded in security and privacy, our vision has always been to build a privacy-first blockchain network to power a decentralized data economy. We have never strayed from our undaunting focus on enabling the use of data and derivative data products, such as AI models, in a manner that embraces and augments the Web3 ethos of transparency, agency, decentralization and governance.

As we launch the Oasis brand refresh guided by community input, we are strengthening our support to developers creating AI dApps that deliver programmable confidentiality to their users. We are proud to have been chosen by one of the top AI-powered protocols in crypto, Ocean Protocol, as the confidential infrastructure of choice for them to develop Ocean Predictoor, a prediction market developed natively on Oasis Sapphire. We are happy to note that Predictoor is approaching an impressive milestone of almost $1 billion in monthly volume after its launch in September 2023.

As dApps run on Oasis technology, they ensure that decentralized AI can work in sync with Smart Privacy as the next generation tech stack. Inclusive and accountable dApps that utilize AI in a decentralized manner can be built leveraging Oasis Sapphire’s smart privacy-preserving features.

In this blog, we will take you on a short journey on how we are realizing our vision for a decentralized AI.

Engineering a more decentralized digital future

New technological developments allow for extending trustless computation provided by mature blockchain platforms into physical infrastructure, traditional data assets and AI. But in the context of Web3, these advancements will not be successful if the Web3 ethos is not upheld through sufficient decentralization.

Oasis can ensure verifiability and confidentiality of arbitrary computation in a decentralized environment. From autonomous AI agents, NFT collections that have a mind of their own, decentralized AI training, oracles, to chain abstraction and defense in depth, the sea of possibilities for privacy-powered AI is endless.

On the Engineering front, Oasis is building upon the proved-and-tested Oasis Sapphire stack by adding additional features that will empower teams building AI dApps to add confidentiality features to protect their users.

ROFL: The Decentralized, Private Infrastructure for AI

As Gen AI and AI-as-a-Service become ubiquitous, there is a large risk of exploits such as the following privilege escalation and cross-tenant attack on inference pipelines. Securing these AI pipelines with confidential computing technologies for both model training, using both open and permissioned data feeds, and inference pipelines is paramount. Confidential computing using Trusted Execution Environments (TEEs) enables training and inference algorithms to run in encrypted memory with runtime integrity, such that their functioning cannot easily be tampered with by any other external cloud service.

Runtime OFf-chain Logic (ROFL) is a framework that adds confidential computing support for off-chain components to runtimes like Oasis Sapphire, enabling not only non-deterministic behavior such as using randomized algorithms and accessing network resources  but also the required confidentiality and runtime integrity for AI pipelines. ROFL allows for off-chain components to seamlessly communicate with the on-chain realm, bringing about full composability across different blockchain platforms and off-chain computation stacks. This enables not only tracking provenance of the data used in AI training, but also provenance of the AI models as derivative data products used in inference pipelines with confidentiality and integrity to the data providers, model providers and, by extension, to AI-as-a-Service providers.

These components are secured by the same Oasis TEEs, the consensus layer and its decentralized validator set, which can transparently run ROFL without the need to install anything besides the Oasis Core nodes and runtime bundles (that node operators currently run). ROFL can be added to any confidential runtime, existing or new, to extend the runtime’s capabilities.

To learn more about how Oasis deployment of ROFL opens up new possibilities for AI, such as how decentralized agents powered by AI can have private “thoughts,” check out the blog written by Oasis Foundation Director, Jernej Kos.

How Oasis Collaborates with AI Teams

Ocean Predictoor — Prediction-Based Trading DApp

Ocean Predictoor is a crowd-sourced, on-chain dApp for making and verifying predictions. Individuals can submit predictions and stake on them, and money is made when they are correct and lost when they aren’t. Traders can buy aggregated accurate predictions in the Predictoor dApp and use them to buy or sell certain assets or contracts.

Using the confidentiality of the Sapphire network, Ocean relies on Oasis technology to secure the execution of its AI-powered on-chain marketplaces. Without this confidentiality, the information collected and shared by parties using the Ocean oracles and AI products would be exposed to anyone in the world. The success of the new Ocean Predictoor dApp has forged a strong partnership between the two teams. Now, both foundations are taking additional steps to expand their efforts.

Currently, Ocean, in collaboration with SingularityNET and FetchAI, has established the Artificial Superintelligence Alliance (ASI) to continue open-source AI research and development. They look to construct a decentralized AI infrastructure that serves as a viable alternative to the centralized AI ecosystems dominated by Big Tech.

deltaDAO — AI Data Management

Recognizing the benefits of Oasis’ parallel runtimes — the separated consensus and compute layers that can replicate confidential computing environments with a shared state — deltaDAO has launched its Pontus-X ecosystem that transforms the way traditional businesses are able to monetize their AI & Data products and helping them transition from Web2 into Web3 where they can fully utilize their digital goods while retaining control over their data.

The launch of Pontus-X improves how different industries can share and monetize their data assets. It breaks down the data silos that exist within the Web2 world to unleash the power of AI & Data. Pontus-X aims to simplify how companies monetize their data while preserving full control and privacy, leveraging the unique capabilities of our network. Through end-to-end data exchange and service monetization, deltaDAO is building a data ecosystem to ensure regulatory clarity and trusted solutions for businesses and users to train and use AI.

Why Confidentiality Matters for AI

AI needs sources of data to train on, and while some data is publicly-available and open to use, much of it is the intellectual property of data providers or individuals. The use of this data may require specific permissions or involve compensation. In addition, workloads may be compute-intensive in a way that blockchain technology on its own cannot provide. Oasis is uniquely positioned to address these needs through its confidentiality features and frameworks.

The Oasis team is developing a new offering, which builds off of the previous success of the consensus layer and Sapphire ParaTime, combining them with verifiable off-chain computation that uses the same underlying cryptography powering the Oasis Network today. Verifiable off-chain capabilities will address both the need for compute intensive workloads in AI, and also the data confidentiality and runtime integrity required when proprietary data is used in AI training. This, we believe, is the cornerstone of decentralized AI.

The Future of Runtimes

TEE runtime extensions can enable complex computations to be performed in an off-chain confidential environment while allowing for the execution integrity to be verified on-chain. This opens up the way towards truly decentralized AI models with both verifiability and confidentiality. For example, AI models trained for facial recognition can securely handle sensitive data while Sapphire smart privacy features ensure that data contributors are compensated for their data and that the AI models are not biased. Furthermore, this enables verification of compliance with data protection laws which is especially relevant in industries such as healthcare, where the privacy and security of patient information needs to be protected. The Oasis decentralized confidential computation platform ensures that AI models operate securely to safeguard sensitive information against unauthorized access.

You can read more about our off-chain confidential computing extension in our ROFL blog. ROFL extends Sapphire, our confidential EVM runtime, and together with OPL, the Oasis Privacy Layer, provides off-chain confidentiality to any dApp on Oasis or any other EVM compatible network.

Support for Data Management

Oasis’ decentralized key management system ensures controlled and auditable data access. This feature enables secure data collaboration across various sectors, including finance, where encrypted data can be analyzed to extract valuable insights without exposing underlying information. The secure environment and decentralized key management allow for the cryptographic validation of origin and integrity of the data provided, combating deep fakes and unauthorized content duplication.

This is just the beginning of what Oasis can do with AI, and the possibilities are nearly endless. If you are looking to build on Oasis and see how decentralized AI and Smart Privacy can be used to their fullest potential, as well as stay updated with relevant hackathons, get in touch here.