Will AI Transform DeFi and Blockchain oracles – Or Create the Next Big Risk for Investors?

Since OpenAI unveiled its flagship ChatGPT product at the end of 2022, nearly every sector is at risk of disruption from the AI revolution. Decentralized Finance (DeFi), itself a disruptor of traditional finance, is quickly seizing the chance to integrate AI. However, in the rush to be the first wave of AI adopters in DeFi, we cannot allow ourselves to compromise on the principles of safety that underpin DeFi and blockchain technology. 

Smart contracts, with their open, permissionless design, provide the ideal playground for AI technologies. In 2024 large language models can be leveraged to interact with the vast DeFi ecosystem and create new layers of value. This has sparked an avalanche of innovation, with self-learning algorithmic traders, on-chain agents executing complex financial strategies, and autonomous algorithms forming organic communities around AI-generated tokens.

However, is AI reliable enough that we can trust it with this amount of value? 

As the AI craze continues, we’re edging closer to a future where AI could take over many DeFi primitives. One of the most talked about use cases is price discovery for financial assets—services typically known as oracle networks, which collectively secure roughly 70 billion dollars. 

The Risks of Accelerating AI Adoption In DeFi

DeFi thrives on experimentation and disruption, values which have it allowed it to grow to over $165B in Total Value Locked. The innovations of DeFi have brought with them new risk vectors, the most notable of which being smart contract exploits. As much as it could add new value to the space, AI could introduce additional risks for developers. 

This makes it crucial to avoid rushing exploratory AI software into production-grade environments. Marcin Kaźmierczak, COO of RedStone, the fastest-growing blockchain oracle, argues:

“There’s no denying that the oracle industry holds significant potential for AI disruption,” says Marcin Kaźmierczak. “However, DeFi shouldn’t rush into adopting AI oracles at one swing. The most critical aspect of oracles is their reliability and security, along with the ability to deliver accurate data. Large Language Models (LLMs) are not at a stage where we can fully trust them to automatically execute oracles flow and safeguard the billions of dollars stored in DeFi. That being said, AI tools have immense value in areas like data monitoring, detecting anomalies, and flagging security concerns for predictable algorightms or humans to act upon—areas that we’re actively exploring and implementing within the RedStone framework.”

Blockchain oracles are designed to provide a trusted source of truth by aggregating and rigorously filtering data from multiple sources, ensuring that DeFi applications receive accurate, reliable inputs. Marcin Kaźmierczak further explains:

“Imagine if we handed over full control of the decision-making process—like detecting and selecting market price sources for a price feed—to a large language model (LLM). Suppose this model identifies a new, highly liquid market that’s been temporarily spun up as part of an attack on a protocol and decides it’s appropriate to predominantly switch to this source for quoting, based on its predefined criteria. The problem is, with oracles, there’s no room for a single mistake. Unlike other systems where an AI model can self-learn and improve after an error, in DeFi, one wrong decision can result in billions of dollars lost, and irreparable damage to a project’s reputation and user’s financial wellbeing.”

Artificial Intelligence and Blockchain: Will AI agents transform DeFi?

For some time now, there has been speculation that AI agents—essentially advanced bots powered by LLMs capable of interacting with smart contracts—will eventually dominate on-chain activity. AI models are continuously improving making it challenging to predict how quickly AI activity with dominate blockchain activity and the degree to which there humans with leverage AI for on-chain activity. Blockchain technology also seeks to be permisionsless allowing the opportunity for autonomous AI models to engage in blockchain activity another interesting topic worth engaging in. 

In recent weeks AI agents have captured the crypto community’s attention. Models like Luna and Terminal of Truth, with autonomous social presences, and recent developments like Coinbase’s ‘Based Agent’ SDK—signal that the convergence of crypto and AI may arrive sooner than expected.

A direct impact of AI agents could be heightened stakes in prediction markets, both in bets on the upcoming 2024 US elections and in future prediction markets. Prediction market giants like Polymarket and Gnosis have been working with AI agents since earlier this year, and AI agents are actively trading in Polymarket’s US election markets. 

But why is the idea of fully autonomous AI models navigating the on-chain economy so exciting? Imagine a “ChatGPT moment,” but for DeFi—a scenario where AI autonomously optimizes your financial resources based on a quick conversation about your risk profile and goals. This could level the playing field for everyone, as deep knowledge of DeFi or finance would no longer be a major differentiator; those decisions could now be handled by unified models. ​​Envision highly profitable strategies, previously reserved for experienced investors, like lending market liquidations, now accessible as subscription-based services provided by on-chain agents. In some ways, this parallels the transformative impact ETFs had on traditional markets by offering straightforward access to diverse assets. However, there are significant risks in granting AI agents such high autonomy, and many aren’t fully aware of them. These highly autonomous models can set their own goals and carry them out as they see fit. We’ve already had a glimpse of this with recent Terminal of Truth actions. Imagine if an AI agent discovers an economic exploit within a DeFi protocol—it will likely pursue the most EV+ scenario for itself, without considering the social consequences or economic harm caused to others. Or, if it identifies a highly extractable MEV opportunity, it could degrade the user experience across the entire on-chain economy. These are delicate issues that need careful consideration before entrusting these models with the future of on-chain finance. 

Beyond Decentralized finance: Trust as the Ultimate Asset

With the abundance of AI computation and its relatively low cost, we are witnessing an exponential rise in online activity that mimics human behavior. This surge has led major social media platforms to battle an avalanche of bot activity. However, oracles could serve as a critical last line of defense against this growing epidemic of fake engagement. Blockchain oracles can function as the web’s filters, screening and validating data before it reaches applications, rejecting artificial inputs generated by bots and malicious actors. In this way, they ensure that applications can once again rely on real, human-generated data.

As a “bot shield,” oracles could be integrated into traditional Web2 platforms like Twitter, potentially solving its long-standing bot problem. Even more intriguingly, they could be natively embedded into Web3 environments like Farcaster, offering decentralized solutions to a decentralized medium. The COO of RedStone explains how this approach could reshape the future of online authenticity and user interaction.

“A Farcaster client like Warpact could automatically route all its traffic through an LLM-powered oracle network, filtering out non-human activity and delivering only authentic, human-generated content to end users. This approach could serve as a reliable method for rewarding genuine, on-chain users in campaigns like airdrops—a solution the industry has long been striving for.”

While this vision is still in the future, it might not be as distant as it seems. More and more teams are tackling the challenge of combating fake social activity by leveraging machine learning algorithms to cluster and categorize different types of behavior. This emerging trend at the intersection of AI and crypto technology demonstrates how innovative solutions can address major obstacles, with the potential to reshape how we authenticate and reward online interactions across both decentralized and traditional platforms.

AI and RedStone’s Decentralized Data Feeds 

In spite of the hype, there are significant risks to handing over vital functions to ai-powered oracles. AI still has limitations, full adoption in DeFi, especially for something as foundational as oracle services, could introduce vulnerabilities that the ecosystem is not yet ready to handle. The balance between AI’s potential and the need for cautious, gradual integration will be crucial to avoid unintended consequences. RedStone is actively researching and testing to realize AI’s full potential in DeFi, unlocking new use cases without compromising the security that has made us a leading oracle provider. The journey is well underway, and we’re excited to keep you updated throughout the process! 

About RedStone’s Blockchain Oracle

RedStone is a Modular Oracle specializing in yield-bearing collateral for lending markets, especially LSTs & LRTs. It offers real-time gas-optimized data feeds across 50+ chains and all rollups. Trusted by Morpho, Venus, ether.fi & more.