Blockchain and Artificial Intelligence (AI): Complementary Technologies That Can Make Each Other Better

There have been three major technology revolutions in modern history: mechanical, electrical, and digital. Today, artificial intelligence (AI) and blockchain are each expanding the frontiers of the digital revolution and have the potential to further complement one another to magnify that expansion. 

As with many emerging technologies, the journey from invention, to use case, to widespread application is not linear—and policymakers have the challenging task of creating regulation that encourages innovation while managing risk.

At its core, blockchain allows users to digitally transfer information in a standardized way, without using an intermediary. This has a number of applications, including being able to instantly transfer value on the internet without the involvement of a third party, in contrast to traditional finance which can require multiple third parties to accomplish the same task. The efficiency, resiliency, transparency, and accessibility granted by the blockchain allow for a number of other uses, including decentralized identification, improved data management, and tokenization of real property.   

AI refers to development of systems that can perform cognitive tasks that have traditionally required human intelligence. With the ability to make predictions and generate new content by analyzing large amounts of data and recognizing patterns, AI has widespread applications to support and augment human-led activities—not just in efficiency and productivity, but advancements in knowledge and innovation. 

This white paper examines how these two digital technologies can each work together to improve the performance of the other. With sound frameworks in place, policymakers and regulators can help foster this symbiotic development and enhance economic, technological, and geopolitical outcomes. 

Advancing and Complementary Technologies

The respective characteristics of blockchain and AI position each technology to address challenges caused by the digital revolution, which connected computer systems and eventually created near infinite amounts of data. Blockchain is uniquely positioned to create digital scarcity, establish better control of standardized personal data, and decentralize online services so they are owned and governed by the communities that use them. For its part, AI is uniquely suited to sift through, interpret, and generate actionable insights from the vast amount of digital information that is available online—especially to the extent that it is standardized and not blocked in data silos.

While blockchain and AI solve many of the problems from the last revolution, they nonetheless create new obstacles. What comes next is a real challenge with far-reaching ramifications: developing a supportive ecosystem, managing negative consequences like fraud and deception, and creating regulatory environments that promote innovation, predictability, and safety.

Countless game changing technologies have followed a similar path. 

The internet is a prime example. Created in 1983 following decades of visionary work, the internet’s full potential took more than a decade to become clear. But recognizing this potential, the Clinton Administration implemented an open framework that allowed domestic innovation to flourish, and the United States to take and hold a leadership role.

Today’s internet, of course, wasn’t the result of any single innovation: it required a complex set of technologies, protocols, and hardware coming together and reinforcing one another. If we had not moved past dial-up modems, for example, today’s digital world—including wifi, streaming services, and video conferencing—would not be possible. 

Similarly, blockchain and AI are two of the most promising technologies of the current digital expansion, and can converge upon one another to ensure mutual amplification.  

The Three A’s: Automation, Analysis, Authentication

 In the coming years and decades, blockchain and AI will intersect and interact repeatedly in order to support the others’ highest aspirations and mitigate associated risks. These areas of intersection include:

1. Automation: Public blockchain networks will enhance AI-driven automation by allowing autonomous agents to hold assets and initiate transactions on those networks.

2. Analysis: AI analysis will be improved by standardized and uniform data generated through blockchain protocols.

3. Authentication: Blockchain will help distinguish between authentic and AI-created content.

1. Automation: Blockchain Advancing AI Integration

Today’s digital world produces mountains of information on a daily basis. As technology advances, so does our ability to process vast amounts of data at speeds that were previously unheard of. But if the processing and analysis of that information includes even one manual component, then the entire information flow will be inevitably slowed. AI removes the need for such manual components, thus fully unleashing this data for optimum utility. 

AI is already touching the lives of millions. For example, Amazon customers benefit from the Amazon Forecast service, which uses AI to parse through historical purchases, weather, holidays, and promotions to ensure that items are moved closer to where the customers are most likely to buy them. Amazon also uses AI to determine which fulfillment centers should be used for any given customer order, analyzing factors like the availability of that item at a given center and its distance from the customer. 

In a not-so-distant future, AI could similarly manage industrial inputs for a factory and make smart, efficient, and automatic reordering decisions. But in order to make these orders, AI will need to be able to pay for them. 

One might assume that giving AI access to a bank account is the answer. But many of the reasons that people don’t like bank accounts—problems with access, moving funds, interacting with customer service—creates significant interoperability challenges for AI. True automation requires settlement guarantees, something a traditional bank account–even one accessed by AI–cannot provide.

That’s because accounts—whose origins date back many millennia—facilitate ownership in a world where assets belong to legally identifiable entities like people and corporations. Today, accounts are created by a third party (like a bank) collecting information about an entity to verify its identity, suitability and related risks— a process known as Know Your Customer. That third party then authenticates the account owner for each transaction.

 Crypto tokens—bearer instruments and stores of value whose provenance is tracked on a blockchain—are a newer form factor that’s better suited for a digital economy. They rely on cryptographic identity, which is universal, and self-authentication with the use of a cryptographic key. Any piece of software or hardware, including an AI agent, can be given a blockchain address that holds token balances and grants the ability to sign transactions. 

This type of automation that leverages open, interconnected economic networks can be further enhanced with the use of smart contracts—sets of conditional rules that are used to effectuate automatic transactions and enforced objectively by the validators of a blockchain network. One concern about giving AI agents the ability to hold and spend money is the possibility of bugs or errors leading to irreversible and costly transactions. But blockchain makes it easy to impose rules on how AI agents transact; rules that are executed by the platform independently of the agent via smart contracts that are deployed on the blockchain network.

2. Analysis: AI Making Use of Standardized Blockchain Data

Today most data not only has many standards, but different counterparties usually keep their own records, leading to reconciliation errors. One of the biggest benefits of any blockchain platform is the standardized data it creates—information that is transparent and trustless, meaning anyone can access it, and no individual participant can alter it. This data can also be verified and shared by many counterparties, thus making it more reliable.

More standardized data has broad potential benefits, and many of the promising blockchain-related applications are focused on how to best process and use that data. However, the technology still needs to be refined to increase adoption. That data is already being put to use by analytics companies and compliance vendors, but there is an opportunity to do more.

 With its ability to analyze vast of amounts of data, AI is well suited to:

• Digest blockchain data for business intelligence, economic analysis, and prevention of illicit activity: AI and data analysis have long been used to check on-chain payments and transactions for suspicious patterns and potential financial crime. Large Language Models (LLMs) in AI systems are exceptionally valuable for fraud detection due to their ability to establish connections between user patterns and profiles, enabling them to efficiently identify patterns that might be out of character. Those algorithms can then be deployed in real-time and become more predictive as they spot patterns and make decisions that separate signal from noise.

For example: Most blockchain analytics and forensic companies deploy AI, including Blocktrace, Arkham, Ciphertrace, Nansen, and Chainalysis. Blocktrace has developed AI to interact with the Bitcoin blockchain, so that users can quickly access specific data points including Bitcoin addresses, transaction dates and values. Instead of having to program in SQL to search the blockchain, users can use a chatbot that’s already trained on the data to find the results they need.

• Improve blockchain security: The bearer nature of digital assets makes breaches and hacks irreversible, and potentially catastrophic. Blockchain platforms and the smart contracts that ride on top of them need to be bulletproof from the start. Given the difficulties of eliminating every bug and foreseeing every attack, AI can be deployed to stress test new solutions before they go live.

Examples: The advantage of historical blockchain data, often accessible from the genesis block or the first block on the chain with technologies like Snowflake, allows for thorough analysis of even the oldest activities. Thanks to the power of AI, no fraudulent activities occurring on the blockchain can escape detection, regardless of how much time has elapsed since their occurrence. This robust combination of AI and blockchain technology significantly enhances the security and reliability of digital payment systems. Companies like OpenZeppelin, Flipside Crypto, and CertiK are already taking preliminary steps towards this future.

In turn, blockchain technology can improve AI through the creation of: 

• Decentralized data markets: AI models are only as good as the data they are trained on. Today, they are trained on troves of public and private data, with little attribution, compensation, or ability for the owners of that data to opt out. In time, blockchain can be used to build decentralized data markets where users opt in to share their data with a model in exchange for compensation, while still preserving their privacy.   Example: Ocean Market’s blockchain-powered decentralized market for data, where users can publish, buy and sell data, empowers developers to use its platform for large language models as it requires minimal preparation. The company is already marketing the ability to tokenize intellectual property as a mechanism to provide compensation from those who use its datasets to train AI models.

3. Authentication: Blockchain Mitigating AI Disinformation Risks

For all its potential benefits, generative AI has the ability to supercharge disinformation in an age where it’s already difficult to tell fact from fake. In the months since AI technology took society by storm, there have been a growing number of instances where AI’s deepfake capabilities were used by a variety of different actors—at times for nefarious purposes. AI also enables impersonation of identities and makes it difficult to tell the difference between humans and bots. We’ve already seen this come to life, from faked images showing the Pentagon was attacked to how AI is being used in the Russia-Ukraine war to spread disinformation.

Blockchain authentication can help solve this problem. One of the key challenges of AI is that its decisions are not transparent – and this is where blockchain can help. Key characteristics that are inherent to the blockchain, such as its transparency, immutability, and ease of access, make it an ideal tool for authenticating the content we see online. One approach is to leverage the blockchain's native use of cryptographic hashes. These are one-way mathematical functions that can take any type of digital data (like text or a photo) as an input and create a verifiable string of fixed length as the output. Storing hashes of important content on-chain creates an immutable record of its original form along with a timestamp. This exact use-case is why the first blockchain was invented all the way back in the 1990s.

Example: IBM is making strides to use blockchain as the “source of truth” for data used to drive specific AI projects. Riley is an IBM Watson AI enabled application which allows the visually impaired to use smartphones to get a description of the areas around them. Given the importance of the accuracy of the data, the creators of Riley are providing a blockchain solution to authenticate information fed into AI systems.

Hashes are effective for creating "digital fingerprints" of content because the same input always yields the same output. Today, we can use a public blockchain like Bitcoin or Ethereum for this purpose. An additional layer or sidechain could be used by content producers to store hashes of their work cheaply, and web browsers could be upgraded with APIs that constantly check the authenticity of photos or text. Going one step further, a content creator, like a news outlet, could cryptographically sign their work with their private key and readers can verify that signed work against the creator’s public key, which is stored on-chain. 

Example: Back in the early 1990s, cryptographers Stuart Haber and Scott Stornetta used an early form of blockchain to timestamp digital documents to verify their authenticity. The product, called AbsoluteProof, acted as a cryptographically secure seal on digital documents. To make it nearly impossible for anyone to backdate timestamps or validate electronic records that were not exact copies of the original, they published the unique hash value of all new seals in the classified section of the New York Times on a weekly basis.

Another path is turning content into NFTs, unique tokens that are stored on the blockchain and can be used to establish the provenance of images, videos, music, and other media. When an NFT is minted, it contains a unique identifier—either a pointer to content stored elsewhere or the content itself—and the initial owner. Minting and transacting NFTs incurs fees, so this approach will likely only be used for the most sensitive content.

And blockchain can do more than authenticate how content was created – it also can track misinformation and how it spreads. Tracking AI-generated content across public blockchains is more feasible than tracking across siloed databases, which ultimately provides a valuable tool for fighting misinformation. This ability will give an upper hand to Web3 versions of services for video, audio, and social media.

There is also an opportunity to create tokenized incentive models for data curation, where users can be rewarded for curating quality content and downvoting inauthentic content.

Finally, blockchain can help identify individuals and tell the difference between bots and real people. The cryptographic identity mentioned earlier is a powerful tool for confirming communication is coming from an expected counterparty and not an AI-driven impersonator.

Going one step further, protocols can be built on top of existing blockchains to help establish “proof of humanity.” One approach is to tie biometric information such as fingerprints or iris scans to an individual’s public key, using newer forms of cryptography (including privacy enhancing technologies) to protect their privacy while still allowing others to confirm they are dealing with an actual human. Another approach is to create a social graph where verified users attest to the humanity of others, using a token to incentivize honest attestations.

Example: World ID is a cryptocurrency project co-founded by OpenAI CEO Sam Altman and backed by Andreesen Horowitz and Coinbase Ventures. With 2 million users behind the program, which creates digital identifications for customers after an iris scan, the project rewards users with a token called Worldcoin. While still in its early stages, World ID would allow users to authenticate and differentiate between AI bots and real people while still preserving their privacy.

The Role of Policy

How these technology revolutions will play out, including their ability to support each other, will depend on pragmatic public policy that encourages responsible growth, provides guidance, and mitigates risks.

While there are important ongoing debates about specific issues, especially those of a technical nature, the best practices for policy are known: create an innovation-friendly culture that allows for technologies to flourish, while setting clear and practical safeguards in specific areas. This is particularly true for foundational technologies like blockchain and AI that have widespread applications. 

Many countries have responded to the rise of crypto and the blockchain with clear-headed pragmatism, understanding this new technology cannot be crammed into existing policy frameworks—many of which pre-date the internet and even the invention of the microchip.  

The consequences of getting policy right are as profound as they are widespread, and the stakes couldn’t be higher right now for both blockchain and AI, and they stand to benefit from each other. The countries that create programmatic policy will win the many benefits of these technologies while providing important guardrails that mitigate their downsides. Those that decline to do so, whether through apathy or hostility, will surrender most of those benefits and be left to manage their risks without the proper tools. 

While we are still in the early innings, we would suggest that policymakers consider four fundamental tenets in their thinking:

1. Clear regulatory guidelines about definitions and classifications

2. Appropriate disclosures and transparency

3. Assurances of proper risk management (particularly when AI is transacting in digital assets)

4. Guidelines around governance and conflicts of interest

As with technology pioneers of the past, most leading blockchain and AI companies want and are strong advocates for reasonable and pragmatic regulation. That’s because we understand that smart policy enables sustainable growth while mitigating risk. The bottom line, like with any transformational technology, is that both blockchain and AI are significantly changing the world we live in. What we want is the opportunity to proactively shape these technologies and the way we interact with them for the better.

The views and opinions expressed herein do not necessarily reflect the views of Coinbase or its employees. This article is intended for informational purposes only, and is not (i) an offer, or solicitation of an offer, to invest in, or to buy or sell, any interests or shares, or to participate in any investment or trading strategy, (ii) intended to provide accounting, legal, or tax advice, or investment recommendations or (iii) an official statement of Coinbase.  Although Coinbase may have financial interests in, or relationships with, some of the companies discussed or referenced herein, such discussion or references should not be viewed as, an endorsement or guarantee of any type by Coinbase.