Verida is enabling the privacy preserving AI tech stack

The Verida Network provides storage infrastructure perfect for AI solutions and the upcoming data connector framework will create a new data economy that benefits end users

This post is part of a Privacy / AI series. See Part 1: Top Three Data Privacy Issues Facing AI Today and Part 2: How web3 and DePIN solves AI’s data privacy problems.

Verida is providing key infrastructure that will underpin the next generation of the privacy preserving AI technology stack. The Verida Network provides private storage, sources of personal data and expandable infrastructure to make this future a reality.

Let’s dive into each of these areas in more detail.

Private storage for personal AI models

The Verida network is designed for storing private, personal data. It is a highly performant, low cost, regulatory compliant solution for storing structured database data for any type of application.

Data stored on the network is protected with a user’s private key, ensuring they are the only account that can request access to, and encrypt their data (unless they provide permission to another account).

This makes the Verida Network ideal for storing private AI models for end users. The network’s high performance (leveraging P2P web socket connections), makes it suitable for high speed read / write applications such as training LLMs.

Source of data for training AI models

We’ve all heard the saying “garbage in, garbage out” when it comes to data. This also applies to training AI models. They are only as good as the data they are fed for training purposes.

The Verida ecosystem provides a broad range of capabilities that make it ideally suited to being a primary source of highly valuable data for training AI models.

Verida has been developing an API data connector framework that enables users to easily connect to existing API’s of centralized platforms and claim their personal data, that they can securely store on the Verida network.

Users on the Verida network will be able to pull health activity data from the likes of Strava and Fitbit. They can pull their private messages from chat platforms, their data from Google and Apple accounts. This can all then be leveraged to train AI models for exclusive use by the user, or that data can be anonymized and contributed to larger training models.

Establishing a data-driven token economy offers a promising avenue for fostering fairness among all stakeholders. Eventually, major tech and data corporations may introduce a token system for service payments, thereby incentivizing users to share their data.

For an example; individuals could leverage their anonymous health data to train AI models for healthcare research and receive token rewards in return. These rewards could then be utilized for subscribing to the service or unlocking premium features, establishing a self-sustaining cycle where data sharing leads to increased service access as a reward. This model fosters a secure and equitable relationship between data contribution and enhanced service access, ensuring that those who contribute more to the ecosystem reap greater benefits in return.

Users could also use their personal data to train AI models designed just for them. Imagine a digital AI assistant guiding you through your life. Suggesting meetup events to attend to improve your career, suggesting a cheaper greener electricity retailer based on your usage, suggesting a better phone plan or simply reminding you of an event you forgot to add to your calendar.

Expandable infrastructure

As touched on in “How web3 and DePIN solves AI’s data privacy problems”, privacy preserving AI will need access to privacy preserving computation to train AI models and respond to user prompts.

Verida is not in the business of providing private decentralized computation, however the Verida identity framework (based on the DID-Core W3C standard) is expandable to connect to this type of Decentralized Physical Infrastructure (DePIN).

Identities on the Verida network can currently be linked to three types of DePIN; Database storage, Private inbox messages, Private notifications. This architecture can easily be extended to support new use cases such as “Private compute” or “Personal AI prompt API”.

With the appropriate partners who support decentralized private compute, there is a very clear pathway to enable personalized, privacy preserving AI leveraging a 100% DePIN technology stack.

This is incredibly exciting, as it will provide a more secure, privacy preserving solution as an alternative to giving all our data to large centralized technology companies.