The Impact of AI + Crypto

Summary: The integration of artificial intelligence (AI) and blockchain is transforming industries by decentralizing infrastructure, enhancing security, and enabling autonomous financial management. On Feb 27, 2025, HTX hosted a Twitter Space on “The Impact of AI & Crypto.” In this AMA, experts from Io.net, Alethea AI, Zark Labs, HTX Research, and Orbit discuss the synergy between AI and crypto, decentralized AI versus centralized models, AI’s role in security and trading, and the future of AI-driven DAOs and metaverse experiences. They explore both opportunities and challenges, emphasizing transparency, decentralization, and responsible AI deployment. Here is a condensed version of the conversation.

Moderator: HTX Host

Speakers: Kate Lane (Io.net, Partnerships Lead), Brent Homesley (Alethea AI, VP of Community), Spiridon Zarkov (Zark Labs, Founder & CEO), Chloe Zheng (HTX Research, Researcher), Ryan McNutt (Orbit, Co-founder & CEO)

Speaker Introductions:

Kate Lane (Io.net): Hey, great to meet everyone. My name is Kate Lane, and I represent Io.net. I lead our Web3 partnerships at Io.net. For those unfamiliar, we’re a decentralized physical infrastructure network that functions like the Airbnb of GPUs. We’ve aggregated underutilized computing power from data centers and individual users worldwide to create the largest decentralized supply of GPUs and CPUs. This network powers leading AI, crypto, and Gen AI companies, offering a decentralized alternative to traditional cloud computing.

Brent Homesley (Alethea AI): Hi, my name is Brent, and I’m the VP of Community at Alethea AI. Alethea AI is a research lab working at the intersection of AI and blockchain since 2019. We were working with OpenAI before ChatGPT was a thing, partnering with them when access to GPT-J was limited. We created the world’s first intelligent NFT (INFT), Alice, trained on Satoshi’s white paper and Alice in Wonderland. She was auctioned off in 2021 for about half a million dollars. We’ve since worked on INFTs for broader use and are now focused on AI agents and their innovations.

Spiridon Zarkov (Zark Labs): Hey everyone, I’m the founder and CEO of Zark Labs. We’re still a new company — only about six to seven months in — but my team and I have a background in the payments industry. We’re building foundational models specifically for blockchain transactions and data. Our goal is to make it easier for people to get into the crypto industry using AI-powered search tools, assist developers in launching applications faster, audit smart contracts, enhance security, and optimize decision-making on-chain.

Chloe Zheng (HTX Research): I lead listing research at HTX, where we evaluate AI projects, including AI meme projects like ACT and Farcon, and framework projects like AI 16.0. We also focus on gaming ecosystems on Solana and BASE. Before joining HTX, I interned at DeepThink, where I learned about AI in the early stages of my career. Our mission at HTX is to support innovative AI-driven projects by assessing their community impact and technological potential.

Ryan McNutt (Orbit): Hey, everyone, I’m Ryan, co-founder and CEO of Orbit. Orbit is essentially a ChatGPT for Web3. We support 117 chains and nearly 200 protocols, allowing users to automate transactions across multiple protocols using AI-powered agents. Our focus is on making Web3 interactions seamless by leveraging AI to streamline on-chain activity. We’re excited about the potential of decentralized AI models and the new possibilities they open up for automation and security in crypto.

Q1: How do AI and blockchain complement each other?

Kate Lane (Io.net): “This convergence has been a long time in the making. AI and blockchain share the common goal of decentralization. In the AI space, we see an increasing need for trust, transparency, and decentralization — elements that blockchain provides inherently. At Io.net, we’re addressing a crucial part of AI’s growth by solving the compute problem. AI models require enormous processing power, and by aggregating underutilized GPUs from individuals and data centers worldwide, we’re creating an accessible, decentralized infrastructure. This removes the dependency on centralized cloud providers like AWS and Google Cloud, making AI computation more affordable and scalable. As AI models become more advanced and integrated into decision-making systems like DAOs and financial markets, blockchain ensures these AI-driven decisions remain auditable, secure, and free from manipulation. I think we’re only scratching the surface of what AI and blockchain can do together.

Brent Homesley (Alethea AI): I’m incredibly bullish on AI and blockchain. Web3 has a rare opportunity to outpace Web2 because of AI’s ability to operate on-chain autonomously. The reason AI and blockchain need each other is simple — AI provides intelligence, while blockchain provides trust and security. In Alethea AI’s work with intelligent NFTs (INFTs), we’ve built AI-powered digital beings that can interact with users, evolve based on blockchain-verified interactions, and execute transactions independently. The ability for AI agents to not just assist but autonomously transact on-chain unlocks a new level of Web3 development. Imagine AI-driven DAOs where decision-making is streamlined by autonomous AI agents analyzing community sentiment and executing governance decisions in real time. We’ve already seen AI-powered trading bots and automated DeFi strategies emerge. This synergy between AI and blockchain will continue to grow, making decentralized systems more efficient and scalable.

Q2: Decentralized AI vs. Centralized AI — What are the challenges?

Brent Homesley (Alethea AI): Just months ago, decentralized compute wasn’t practical. Now, thanks to projects like Io.net, it’s a reality. The next challenge is making decentralized AI match centralized performance, but open-source innovation is quickly closing that gap. Compute availability used to be a major barrier, but now we’re seeing more accessible networks providing GPU power at competitive rates. The next big challenge is ensuring that decentralized AI models can be properly trained and optimized without relying on centralized infrastructure.

One of the biggest concerns with decentralized AI is security. With centralized AI, companies can enforce strict protocols to prevent data leaks and ensure compliance with regulations. However, in a decentralized model, there must be mechanisms to protect against malicious actors who may try to manipulate AI models by injecting biased or false data. This is an area that needs improvement before decentralized AI can truly compete with big tech.”

Ryan McNutt (Orbit): Decentralized AI is exciting, but scaling it is tough. We’re seeing open-source models rise to compete with centralized AI, and infrastructure is evolving to support wider adoption. Incentives need to be in place to encourage resource sharing for better AI models. One issue we’re facing is the complexity of getting decentralized AI to function at the same level as centralized AI while maintaining efficiency. There are trade-offs in terms of cost, accessibility, and speed, but as the technology advances, we’re going to see more adoption.

Q3: AI’s Role in Crypto Trading & Security

Spiridon Zarkov (Zark Labs): We’re shifting from rule-based security to AI-driven fraud detection. Traditional systems use hundreds of models, but AI allows real-time analysis to catch fraud before it happens. AI enhances risk detection, making blockchain transactions safer. AI models trained on blockchain data can predict fraudulent behavior by analyzing patterns in real-time transactions, preventing scams before they take effect.

For example, in traditional finance, fraud detection relies on rules-based systems that flag unusual transactions. AI goes beyond that by learning from patterns and behaviors, making it better at identifying new and evolving fraud tactics. This is a crucial innovation in the crypto space, where scams and exploits are prevalent.

Chloe Zheng (HTX Research): AI should monitor trading but not control it entirely. If an AI bot makes a costly mistake, who’s responsible? Security frameworks like ZK and TEE ensure that AI aids security without introducing new risks. AI can be extremely useful for fraud prevention, but it should work alongside human oversight rather than replacing it entirely.

Ryan McNutt (Orbit): Our trading agents work within strict parameters — they analyze large datasets but operate only within predefined rules. Unlike humans, AI isn’t emotional, making trading more efficient and less prone to irrational decisions. AI can help detect market anomalies and prevent manipulation, but it should be used cautiously to avoid unintended consequences.

Q4: Can Decentralized Compute Networks Compete with AWS & Google Cloud?

Kate Lane (Io.net): Yes! Io.net aggregates underutilized GPUs, cutting compute costs by up to 90%. There’s no vendor lock-in — you can spin up clusters in minutes. As AI demand rises, decentralized networks will challenge traditional cloud providers. This means startups and independent developers can train models without the massive costs associated with traditional cloud services.

Brent Homesley (Alethea AI): For Web3, decentralized compute is a game-changer. AI agents will soon autonomously purchase compute as needed. This reduces costs and makes compute more accessible, supporting decentralized AI growth. By making computing resources more available to developers, we remove one of the biggest barriers to AI adoption in Web3.

Q5: AI-Driven DAOs & Autonomous Agents — Can AI Manage Decentralized Communities?

Kate Lane (Io.net): AI will improve efficiency — helping with decision-making and governance — but full autonomy isn’t realistic yet. Treasury management and major strategic decisions still require human oversight. AI can provide recommendations, but people must remain the final decision-makers.

Ryan McNutt (Orbit): I believe DAOs will become fully autonomous over time. AI is already outperforming humans in decision-making. We’re heading toward AI-driven governance, where DAOs can operate with minimal human intervention. The key is designing AI systems that reflect community values while minimizing risks.

Q6: How does Io.net plan to compete with centralized cloud giants in offering cost-effective AI compute power for model training?

Kate Lane (Io.net): In addition to what I shared earlier, we’ve tackled the supply issue, but the next step is differentiating ourselves from hyperscalers. A big part of that is building developer tools to simplify AI model training and deployment. That’s why we launched Io Intelligence — a hub of developer tools designed to remove complexity and make GPU networks more accessible.

One of the first offerings is our inference-as-a-service tool, which allows anyone to query top open-source models through an API, essentially for free. This reduces the technical barriers and cost of AI integration. Now that we’ve solved compute availability, our next challenge is streamlining the developer experience, making decentralized compute as easy to use as AWS but with significantly lower costs and more flexibility.

Q7: AI-driven trading strategies can be powerful but also risky. How does Orbit mitigate AI biases and unexpected market behaviors?

Ryan McNutt (Orbit): Risk in AI-driven trading comes down to data. Our trading agents operate purely on data-driven insights, pulling from 50+ different sources, including technical indicators and even social sentiment from platforms like Twitter.

We mitigate bias by ensuring that agents only receive filtered and pre-processed data, preventing any single dataset from skewing decisions. Additionally, the AI isn’t given unrestricted control — it operates within a strict subset of functions, ensuring trades are executed based on predefined, optimized parameters.

To further improve accuracy, our system supports both user-defined data inputs and agent-to-agent communication, allowing AI agents to collaborate for more complex decision-making. By limiting an agent’s scope and ensuring high-quality data ingestion, we reduce the likelihood of erratic or biased trading behaviors.

Q8: With the popularity of AI projects, how does HTX cater to this demand?

Chloe Zheng (HTX Research): At HTX, we evaluate both the community strength and technological foundation of AI projects before listing them. Since Q3 2024, we’ve listed over 10 AI-related token projects. Our priority is ensuring that these projects have strong real-user engagement, not just inflated metrics.

For instance, we’ve listed GOAT from Project Truth Terminal; this AI agent launched the token with a $50,000 donation from the CEO of A16Z. When evaluating AI projects, we look for strong communities and real users versus fake ones. The AI16Z project and the ELISA framework were initially criticized for having no value, but we ultimately listed &AI16Z/ELISA for its strong community as well.

We’re also tracking virtual projects and have listed AIXBT and LUNA, which we believe have significant value due to their ability to attract attention. We care about creating a “wealth effect” for our users to make profits on our exchange. For example, ACT experienced a 50-fold increase after listing.

Our listing team personally tests projects to filter out low-value or fraudulent tokens. We have tested our previously listed DeFi projects GRIFFIN and NEAR to ensure they’re useful and not fake projects. We’re also closely monitoring developments in the BitTensor TAO update, and new models like BitTensor’s subnet token economy. AI projects that demonstrate real-world application, innovation, and community impact have a strong chance of being listed at HTX.

Q9: As the former Head of AI at Visa, how does AI balance security and user convenience in payments?

Spiridon Zarkov (Zark Labs): One of the biggest challenges in AI for payments is minimizing false positives while maintaining strong security. AI models must continuously learn from real-world transactions to improve fraud detection accuracy.

A common example is international travel — if someone makes a purchase in Singapore and, minutes later, a similar transaction occurs in London, old models would immediately flag it as fraud. But as we gathered more data from banks issuing multiple cards, we trained models to recognize legitimate spending behaviors.

In crypto, AI fraud detection is even more critical because transactions are irreversible. Unlike traditional finance, where banks can refund disputed transactions, crypto users don’t have that safety net. AI needs to work proactively, identifying risky transactions before they occur, providing real-time risk scores, and notifying users of potential threats. We need to develop even more sophisticated AI-driven security frameworks to match crypto’s unique challenges.

Q10: AI-Powered Metaverse: What role does AI play in shaping interactive experiences within the metaverse, and how can blockchain ensure authenticity?

Brent Homesley (Alethea AI): AI is transforming the metaverse by making interactions more dynamic. One clear example is AI-driven NPCs — these aren’t just scripted bots but intelligent, evolving entities that interact naturally with users.

At Alethea, we’ve seen our community create AI-powered companions, assistants, and storytelling agents. These agents bring new forms of entertainment and engagement into the metaverse. With blockchain, we ensure authenticity by tying AI-generated assets and interactions to on-chain records, proving their provenance and ownership. As AI-driven virtual worlds evolve, we’ll see deeper, more immersive digital experiences that remain verifiable through blockchain technology.

Closing Thoughts:

AI and blockchain are revolutionizing industries. While challenges like compute power, security, and governance persist, open-source innovation is rapidly advancing decentralized AI. The synergy between AI and blockchain is shaping the future, making it more secure, efficient, and autonomous. As adoption grows, we’ll see more applications in governance, security, trading, and decentralized computing, pushing the boundaries of what’s possible in AI and crypto.