Blockchain Technology May Be About to Find Its Killer Use Case: AI Accountability

With the rapid rise of generative AI models capable of creating texts, pictures, and even videos virtually indistinguishable from reality, policymakers globally are on the back foot.

Generative AI models are capable of creating texts, pictures, and even videos that are virtually indistinguishable from reality. This technology has the potential to revolutionize many industries, but it also poses new challenges for policymakers.

One of the biggest challenges is ensuring that generative AI models are used responsibly. For example, if a deepfake of a politician admitting to a heinous crime goes viral and costs them the election, how do we discover who did it? Or if an AI model developed by a trading house crashes a national economy and causes financial instability, how do we hold the developers accountable?

To address these challenges, some policymakers are considering using blockchain technology to create an immutable evidence trail for generative AI models. This would allow regulators to track who created a particular model, what data was used to train it, and how it was used.

An immutable evidence trail would make it much easier to enforce regulations on generative AI models. It would also naturally disincentivize bad behavior in the first place. After all, people are less inclined to steal when they know a CCTV is watching them.

However, this doesn’t have to lead to a world without privacy; quite the opposite. By using privacy-preserving techniques, such as zk-SNARKs, we can keep sensitive data confidential while still allowing regulators to verify its authenticity.

At the recent London Blockchain Conference, FICO’s Scott Zoldi explained how the data analytics company uses a private blockchain to create accountability in AI. It stores the evidence trail on a private blockchain, which regulators can access as needed.

However, private blockchains have their limitations—one being that they are essentially databases controlled by the same entities regulators may be trying to investigate.

The solution is scalable public blockchains like BSV. These allow for sensitive data to be stored in private overlay networks while being linked back to cryptographic hashes of the data on a public chain. In this way, regulators can verify that data in the private networks has not been altered since its creation. Thus, we have an immutable evidence trail with sensitive data kept away from prying eyes but linked directly back to a public record that cannot be tampered with.

Applied to the development of AI models, this would allow for data about the scientists and engineers, as well as any intellectual property and other confidential information about the models themselves, to be kept private while linking them back to a time-stamped record on a public, immutable blockchain that verifies what the sequestered data says.

Is this pie-in-the-sky hypothetical, or does it have any basis in reality? As a matter of fact, Sentinel Node and Trace App, both blockchain apps developed in conjunction with IBM (NYSE:IBM), are already live and utilizing precisely what is described above.

It’s only with the unanticipated explosion of generative AI models that the importance of such a technology is finally being realized. Of course, for it to work, the blockchain will need to be unboundedly scalable, and policymakers worldwide will have to recognize its potential as a solution and have the political will to mandate its use.