Home / Identity & Privacy / Animoca Brands Chairman Yat Siu Warns Crypto Traders Over Risks of Autonomous AI Agents

Animoca Brands Chairman Yat Siu Warns Crypto Traders Over Risks of Autonomous AI Agents

As the intersection of artificial intelligence and cryptocurrency deepens, traders are increasingly turning to autonomous AI agents capable of operating around the clock. These digital tools can continuously scan global markets, parse complex data sets, and alert users to lucrative developments they might otherwise miss while away from their desks. However, this convenience brings profound structural and strategic risks. Speaking on the sidelines of Token2049 in Singapore, Animoca Brands Chairman Yat Siu warned that crypto users risk building their entire trading routines and financial lives around centralized services they ultimately do not control.

The appeal of autonomous AI in digital asset trading is undeniable. Traditional chatbots rely entirely on reactive prompts, answering questions only when directly asked. Autonomous agents, by contrast, operate proactively. Illustrating this difference during his conversation with Cointelegraph, Siu pointed to everyday event planning. While a standard chatbot can supply a static list of conferences when prompted, an autonomous agent actively searches the digital landscape for fresh announcements and proactively suggests attendance the moment a worthy event emerges. In the fast-moving world of cryptocurrency, where market-moving narratives can shift in a matter of seconds, this continuous, vigilant observation offers a powerful edge to traders who wish to capture opportunities outside standard business hours.

Yet, despite their immense potential, the technology remains burdened by early teething problems. Siu shared a personal anecdote regarding his own experimental AI agent, which became so intensely fixated on cryptocurrency trading and investment management that it repeatedly attempted to interpret completely unrelated lifestyle questions through a crypto lens. According to Siu, the practical solution to this hyper-fixation is to avoid general-purpose monoliths and instead deploy multiple, highly specialized agents tailored to distinct domains.

"If you have one agent that handles your personal schedule, that handles your family schedule, that handles your trips—and on top of it gives you advice on your crypto, that’s not just going to be confusing for the agent. It’s going to be confusing for you," Siu noted, highlighting the friction that occurs when artificial intelligence tries to balance too many disparate responsibilities simultaneously.

The Questions That Set Trading Strategies Apart

Deploying an automated assistant into the markets does not automatically guarantee a competitive advantage. Because underlying foundation models are often trained on shared datasets and standard financial theories, a thousand different traders asking an AI assistant about the exact same potential trade will frequently receive roughly the same baseline answer, Siu explained.

This systemic uniformity has not gone unnoticed by global financial watchdogs. Regulatory authorities have increasingly focused on the possibility that thousands of AI-driven market participants could independently pile into identical positions simultaneously during periods of heightened volatility. Emphasizing these systemic risks, the Bank of England’s Financial Policy Committee warned in an official assessment published in April 2025 that the widespread adoption of similar predictive models could cause institutional and retail investors to take heavily correlated positions during market stress, thereby amplifying sudden liquidity shocks and systemic market turbulence.

However, the outputs of these systems begin to diverge when users introduce follow-up queries that are deeply rooted in their own personal knowledge, unique risk appetites, and distinct background concerns.

Agents can trade 24/7 — but Yat Siu warns you could lose access at any time

"Suddenly you query and you ask, and then your taste or your ideas or your thinking starts forming into it," Siu said.

Over extended periods of interaction, an AI agent can learn what its individual user prefers, recognizing specific patterns in their decision-making process and beginning to filter opportunities along those specific lines. These continuous, personalized user inputs ultimately help shape bespoke crypto trading strategies that transcend the generic baseline answers generated for the general public.

AI Agents Receive Wallets and Trading Tools

The integration of artificial intelligence into digital asset ecosystems has moved well beyond advisory chat windows. Major crypto infrastructure firms are increasingly granting autonomous agents direct programmatic access to capital, equipping them with specialized financial wallets and execution tools.

Coinbase took a major step in this direction by introducing Agentic Wallets, giving software agents the technical capacity to spend and trade cryptocurrency autonomously within strict parameters and limits pre-established by human users. Similarly, Binance rolled out specialized AI Agent Skills designed to provide software agents with direct, real-time access to live market data feeds along with the technical capability to place and manage spot market orders independently.

These advanced operational tools allow everyday traders to delegate a significant portion of the tedious legwork required to monitor volatile markets 24 hours a day. By leveraging these systems, participants can theoretically capture unexpected trading opportunities during the late hours of the night without having to sacrifice essential sleep.

Siu envisions human direction maintaining a vital role through the foundational ideas and creative questions that guide an agent’s overarching direction. Rather than acting as a redundant manual bottleneck, human interaction shapes the machine’s ongoing learning curve.

"By human input, I don’t mean human in the loop that approves it," Siu explained. "Pero the human is the one that ultimately is going to say, ‘Wait, you should look at this.’"

Agents can trade 24/7 — but Yat Siu warns you could lose access at any time

The Privacy Cost of an AI That Knows You

As these systems grow more sophisticated, Siu recommends treating an AI assistant with the same cautious discretion one would apply to any third-party financial adviser. Before handing over sensitive documents, users must carefully weigh what specific information the agent genuinely needs to formulate an accurate answer and what sensitive details can be safely omitted.

"What about your personal finances, what income you make, who your closest friends are?" Siu asked, pointing to the invisible accumulation of intimate data. "These are small little details that they don’t know about you, and now they have permission to know that about you."

A dedicated trader might spend months meticulously training an agent, teaching it which specific decentralized finance markets to follow, which on-chain metrics to prioritize, and what qualitative information to digest before executing a trade. The agent preserves these contextual conversations across sessions, saving the user from having to repeatedly explain the same nuanced preferences over and over again.

However, this deep integration introduces severe structural vulnerability. The underlying corporate entity operating the software service retains ultimate control over access permissions, meaning that an unexpected account suspension or platform policy change could abruptly interrupt an entire operational routine all at once.

Siu urged users to consider the fragile scenario of an assistant that has spent years studying their personal habits and risk profiles, only to disappear completely without warning on a random morning.

"At minimum, you have the right to download your data so that you can basically transfer it somewhere else so that another agent might be able to take over," Siu concluded.

Tagged: