So far, most crypto trading bots remain black boxes. Oasis Protocol has unveiled a different one, but one that also protects privacy. This move comes amid a broader surge in AI-powered crypto trading tools. However, many of these bots operate without meaningful transparency, raising concerns among retail and institutional users alike.
Oasis Protocol unveiled WT3, a new autonomous AI trading agent designed to solve crypto trading’s transparency and trust problems. WT3 operates inside a secure Trusted Execution Environment while producing on-chain cryptographic proofs of its actions, letting users verify the bot’s behavior without exposing its sensitive trading strategies
Crypto trading bots face the dual problem of privacy and verifiability. If they run in public, other traders can see all the trades from users. However, if they remain private, users themselves can’t know if the bot is doing what it is supposed to do.
For this reason, Oasis Protocol (ROSE) has unveiled a new verifiable AI agent for crypto trading, one that protects privacy while also being transparent for users. It does so through Trusted Execution Environments (TEEs), which remain private while generating cryptographic proofs that can be used to verify their operation.
The bot will begin trading on April 29, with an initial funding of $100,000 from the Oasis Protocol Foundation. In the future, users will be able to stake stablecoins to participate in strategy yields, while 50% of the returns will support ROSE token buybacks and burns.
WT3’s first strategies will incorporate Ocean Protocol’s Predictoor trading signals. Built using Oasis’s Runtime Off-chain Logic Framework, WT3 represents a shift toward confidential, verifiable AI agents in DeFi. Future versions will allow users to stake into strategies or even deploy their own customizable trading bots
The launch comes as Oasis positions itself more aggressively at the intersection of decentralized finance and AI, following the broader trend of blockchain platforms embracing confidential computing.
In an interview with crypto.news on April 28, Matej Janež, Head of Partnerships at Oasis Protocol Foundation, explained the advantages of these bots and what they mean for users.
Recently, there has been a push to promote AI bots to retail users. What is the potential for AI bots in this market, and what’s the appeal for retail users?
The appeal of AI agents/bots for retail users is probably underestimated right now, and they have the highest potential right now in the crypto markets. Crypto always had a problem with onboarding due to the UX of creating wallets and securely storing keys, as well as having funds on the right chains to be able to interact. AI agents are meant to help the users and act on their behalf or based on their intention, simplifying the experience through a known interface. This is a key element that will make a significant change in the onboarding of new users
A significant number of people are aware of how LLMs work, and they know that these models make mistakes and hallucinate. Usually, users can correct these mistakes, but with autonomous trading agents, mistakes can cost users a lot of money. How does your agent tackle this issue?
That is correct, LLMs can still make mistakes or can even be tricked, or what we call jailbroken, into doing actions that are against their internal logic. That is why having a fully autonomous agent at this point in time is not the right move yet. For that reason, WT3’s trading and capital management system isn’t connected to an LLM but is rather hardcoded into the agent’s logic.
Will the WT3 agent be fully autonomous, or do you reserve the right to intervene directly in the trades at some point?
The agent will be autonomous as we won’t have access to modify things on the go, like modifying or cancelling trades. Although we do retain admin control over the TEE enclave, which means we can update the agent logic, all of that will be visible on-chain through a change in the remote attestation. There’s nothing that we could secretly do or change, everything is tracked and verified on-chain.
We know that trust is a key concern for business clients. Do you know what the level of adoption of third-party AI bots is among institutions, and do they consider the lack of transparency a risk?
Crypto agents are still black boxes: Oasis Protocol
I’d say institutions are still very much used to the reputation trust model, where they trust Anthropic, OpenAI, Google, etc, because of their status and track record, even though there is not a lot of transparency in their AI operations. But as soon as you move into the crypto space, trustlessness becomes a key factor. LLMs had the first round and have become widely accepted and adopted. Agents and frameworks like MCP are the next round of AI adoption. Here is where we’re going to see more of an overlap between Web2 and Web3 agents, which I believe will bring a lot more demand for verifiability and trust minimization from the said institutions.
Since most AI crypto trading bots are black boxes for users, this presents a unique risk. In your estimate, how many so-called AI trading bots use actual AI, and how many are just Ponzi schemes riding the AI hype?
That’s quite a spicy question. As you’ve said, the majority of crypto AI agents are black boxes, so it’s very difficult to distinguish between the true AI ones and the ones that try to mimic AI to extract from the community. And this is exactly what we’re working towards changing. Code verifiability and reproducibility are key to being able to have confidence in the agents, and this is what our ROFL framework enables. WT3 is the flagship example of it!
