The crypto industry has long been a hotbed of hype and innovation, with promoters constantly touting the next big thing. From supply chain logistics to digital art, the promises have been as diverse as they are ambitious. But in the realm of artificial intelligence (AI), the crypto community seems to be particularly enthusiastic, with some even claiming that blockchain technology is the future of AI. However, a recent report from researchers at Yale, Harvard, Princeton, and other top universities paints a different picture, suggesting that crypto has limited utility for solving issues related to payments and trust in AI.
The report, known as the Crypto x AI Survey, examines the potential overlaps between crypto and AI. While it acknowledges some conceptual possibilities, such as stablecoins and micropayment systems enabling autonomous agents to handle payments for data samples or compute access, it also highlights the lack of significant traction in the payments sector. The survey concludes that AI and crypto are still in the very early stages of meaningful integration, with decentralized governance ideas like DAOs for AI development yet to see real adoption in the mainstream AI community.
One of the most intriguing aspects of the report is its examination of the other side of the equation: whether AI can help crypto. Here, the researchers identify several areas with more immediate applications, particularly in security and systems design. Graph neural networks and other machine learning methods have demonstrated effectiveness at detecting fraudulent transactions or anomalies by examining blockchain data patterns, often outperforming simpler approaches when sufficient training data is available. Large language models can proactively scan smart contract code for vulnerabilities.
However, the report also flags corresponding risks, including the potential for AI systems to power more effective attacks on decentralized protocols or to create unpredictable autonomous agents that interact with smart contracts. AI has become a growing threat in crypto, with one early blockchain security pioneer warning users to stay away from DeFi for now because AI agents could hunt for bugs in smart contracts and use them to drain funds. Coding agents are superhuman at finding vulnerabilities, and smart contract security is too asymmetric: defenders need to fix every bug while attackers need just one exploit to steal funds.
Despite these risks, there are also opportunities for AI to be used defensively by those looking to secure crypto protocols. For example, AI can be used to proactively scan smart contract code for vulnerabilities, as was the case with the recent bug found in Zcash that could have allowed an attacker to print cryptocurrency out of thin air. However, the privacy features of Zcash make it difficult to confirm whether or not this particular bug had been exploited before an associated patch was released.
Two of the most well-known crypto projects associated with AI are Bittensor and the Sam Altman-backed Worldcoin. Bittensor is focused on creating a decentralized network where participants contribute machine learning models and other AI-related services in exchange for token rewards, while Worldcoin has been focused on solving the problem of Sybil attacks on the internet more generally, which are expected to worsen with the advent of AI. Worldcoin also has its own associated cryptocurrency, and its solution for Sybil attacks is for everyone in the world to have their eyeballs scanned with their dystopian orb hardware (seriously). On Monday, Business Insider reported layoffs at the eyeball-scanning startup.
Projects like Bittensor and Worldcoin have an obvious incentive to hype the supposed overlap between crypto and AI to boost their own tokens. But bitcoin advocates and stablecoin issuers have also gotten in on the act, touting their preferred digital currencies as essential infrastructure for the coming AI agent economy. USDC stablecoin issuer Circle has been talking about how AI agents will need USDC for payments on a more regular basis and has built out entire infrastructure platforms for the growth it expects to see in this area. On X, bitcoin-focused media outlet TFTC previously posted, '[AI agents] are choosing Bitcoin over traditional payment systems for a simple reason: it’s easier to spin up a Lightning wallet than set up a Stripe account.'
However, there are also plenty of longtime bitcoin-only users (often referred to as bitcoin maximalists) who have seen this story play out before and are skeptical of the claims around crypto and AI, even when it comes to the supposed utility of bitcoin. According to this recent research, it appears AI is currently more likely to assist crypto than the other way around. Personally, I think this is a fascinating development, as it suggests that the relationship between crypto and AI is more symbiotic than previously thought. It also raises a deeper question: what will happen when AI becomes more advanced and begins to develop its own goals and motivations? Will it seek to protect and enhance the crypto ecosystem, or will it turn against it?
