April 26th, 2026

AI trading agents are only as trustworthy as their data

Key points:

AI agents now pose a greater systemic risk to crypto than traditional hackers or fraud.
Markets are vulnerable because attackers can easily poison the news data that AI agents ingest.
AI often follows patterns without understanding context, leading to immediate and highly amplified market errors.
Minimal capital is needed to trigger a crash by seeding false narratives across social media.
Maintaining human oversight is the most vital safeguard against rapid and synchronized algorithmic market failures.

Imagine a major crypto exchange declaring insolvency out of the blue. In the past, hackers or fraud caused wipeouts worth billions of dollars, but today? AI could just as easily be the culprit.

With AI agents that can autonomously trade on cryptocurrency exchanges being pushed by various players in the industry, agents causing a crypto crash is a plausible scenario.

Simply put, if an AI agent is designed to make trades based on market information – including news articles or social media posts – it would be relatively easy to “poison” those sources with false narratives. This could trigger a wave of automated selling from agents that couldn’t distinguish the rumor from reality, which could then crash a coin or a whole market.

While no such attack has happened yet, the conditions for one already exist. The question is no longer if an AI-driven financial crisis will occur, but when – and, more unsettlingly, how little capital it might take to trigger one.

In my work as an advisor to Web3 companies and government organizations, I have watched the narrative around AI in crypto shift from cautious optimism to uncritical adoption.

Today, 45.7% of platform interactions on Binance are system-triggered rather than user-initiated, which means they are carried out by a computer, not a human. That share is only growing, and every percentage point represents a wider attack surface for anyone looking to exploit these agents.

How AI trading agents work

While AI trading agents are designed to bring efficiency, they are also highly vulnerable. The combination of autonomous agents, high-frequency trading infrastructure, and an information ecosystem saturated with synthetic media has created a perfect storm for potential attacks.

At a basic level, these agents ingest market data – price movements, order books, news, and social sentiment – and use machine learning models to identify patterns or signals that inform trading decisions. Once certain conditions are met, they execute trades automatically, often at high speed and without human intervention.

However, recent research underscores how fragile these agents are in ways that should alarm anyone using them.

A study released in February tested 13 AI trading models using distorted or misleading market data. Most didn’t adapt at all, and their performance barely changed, suggesting they were just following fixed strategies rather than reacting to new signals.

When false signals were introduced, some models saw sharp drops in performance, showing how easily they could be thrown off by bad information.

The study also identified what it calls a “competence mirage”: models that identified the correct trading strategy but got the underlying numbers wrong. Knowing what to do and being able to execute it accurately are, it turns out, very different things.

This serves as a reminder that AI agents aren’t sophisticated market participants but pattern-matching engines operating on the data they are fed. When that data is poisoned through coordinated fake news or purchased synthetic datasets, the reaction is immediate and amplified.

Plan of attack

How would such an attack on crypto trading agents work in practice?

An attacker wouldn’t need large amounts of capital to influence the flow of information that trading systems respond to. That could mean seeding false narratives across news outlets, social media, or data feeds using trigger phrases like “liquidity crisis” or “regulatory crackdown,” prompting the agents to react as if the threat were real.

This isn’t purely theoretical, as false information has moved markets before. When the Associated Press Twitter account was hacked in 2013, a single fake tweet briefly wiped billions off the S&P 500.

Events like the 2010 Flash Crash have also shown how automated trading can amplify shocks at speed. In crypto markets, where sentiment already drives volatility, the bar to trigger a cascade may be even lower.

A relatively well-funded actor could seed false narratives across news feeds, coordinate bot networks to amplify them, and target the data sources that trading systems rely on. Normally, it takes hundreds of millions to move markets, but not in this case.

Protection

There are existing safeguards that can help mitigate these risks, like trading halts or AI-driven fraud detection. Traditional financial markets have mechanisms to halt trading during extreme volatility.

However, these frameworks were built with human behavior in mind and often fail to account for automated systems. As crypto markets operate 24/7 with fewer trading halts, there are a lot more opportunities for attacks.

Others suggest AI will eventually learn to detect manipulation. But research from HEC Paris notes that AI excels at short-term pattern recognition but fails at long-term contextual understanding.

When multiple AI agents rely on similar models and react to identical signals, they tend to make the same decisions at the same time. If those signals are wrong, the mistake spreads across the market, and at the speed of modern trading, that can quickly turn into a wave of synchronized selling.

As with much in AI, keeping a human in the loop may be the most effective safeguard.

The human layer in trading – analysts, compliance officers, and risk managers – shouldn’t disappear but evolve. Their role should be to question information, verify whether news is real, assess where data comes from, and apply judgment that AI lacks.

It may seem like friction to have humans involved. But in a system where speed is the vulnerability, friction is the point.

## What this means for industry players

For founders and investors operating in the crypto trading space, they shouldn’t treat the manipulation of agents as a theoretical risk.

The founders building AI trading infrastructure must position resilience as a value proposition. If they can build systems that can withstand poisoned data, use diverse data sources, and create transparent AI decision pathways, their solutions will stand out.

Meanwhile, investors backing such platforms should look closely at their “human-in-the-loop” protocols. Does the startup rely on fully autonomous execution, or is there mandatory human oversight for critical decisions?

The latter is a safer bet, as the risk of liability in a flash crash scenario driven by an agent’s error is massive.

The convergence of AI and financial products in both crypto and traditional finance is inevitable, but its trajectory is not predetermined. We can choose to build systems that are resilient, transparent, and human-centric, or we can sleepwalk into a future where a few lines of poisoned code cause huge losses.

The choice is ours, but the window for action is closing.

Source: https://www.techinasia.com/ai-trading-agents-trustworthy-data

Anndy Lian

Anndy Lian is an early blockchain adopter and experienced serial entrepreneur who is known for his work in the government sector. He is a best selling book author- “NFT: From Zero to Hero” and “Blockchain Revolution 2030”.

Currently, he is appointed as the Chief Digital Advisor at Mongolia Productivity Organization, championing national digitization. Prior to his current appointments, he was the Chairman of BigONE Exchange, a global top 30 ranked crypto spot exchange and was also the Advisory Board Member for Hyundai DAC, the blockchain arm of South Korea’s largest car manufacturer Hyundai Motor Group. Lian played a pivotal role as the Blockchain Advisor for Asian Productivity Organisation (APO), an intergovernmental organization committed to improving productivity in the Asia-Pacific region.

An avid supporter of incubating start-ups, Anndy has also been a private investor for the past eight years. With a growth investment mindset, Anndy strategically demonstrates this in the companies he chooses to be involved with. He believes that what he is doing through blockchain technology currently will revolutionise and redefine traditional businesses. He also believes that the blockchain industry has to be “redecentralised”.

www.anndy.com

j j j

March 30th, 2026

China Can’t Export Electricity, So It Did Something Smarter: The AI Token Revolution Explained

China’s electricity cannot cross its borders, but Chinese tokens are already sold globally. These two phenomena are essentially the same thing. Tokens are China’s true electricity export. I know this concept may not have fully clicked yet, but every sentence I share is backed by data.

China generates 10 trillion kilowatt-hours of electricity annually, surpassing the EU, Russia, India, and Japan combined. This is not because China lacks the desire to sell. It is physically impossible. Electricity cannot be stored or loaded onto ships. Extending high-voltage transmission lines across national borders involves negotiations that can drag on for a decade. It is like holding the world’s largest gold mine where the gold is too heavy to transport, leaving it piled up in your own backyard.

Tokens have shattered this bottleneck.

First, let us clarify what a token represents. When you converse with an AI like DeepSeek, every character and line of code it returns consists of tokens. On the surface, they appear as text or dialogue. Fundamentally, they are digitally encapsulated electrical energy. If you doubt this, consider the math. In the cost structure of AI inference, electricity plus compute depreciation together account for a staggering 80% to 90%. In other words, nearly 90 cents of every dollar spent on a token effectively pays for electricity.

A token is a compressed packet of electrical energy, representing the final product refined from China’s northwestern green electricity through GPU computation.

So how does this relate to exports? When a Silicon Valley developer sits at their computer and calls a Chinese large language model API, data instantly traverses undersea fiber-optic cables to reach computing centers in Ningxia or Inner Mongolia. Thousands of GPUs roar to life, consuming China’s cheapest northwestern green power to perform logical inference. They return the result to a screen in San Francisco within seconds. Throughout this entire process, not a drop of oil was burned, and not a single power cable crossed a border. The value of Chinese electricity has already been delivered across borders via tokens. This is dimensional warfare involving zero physical output, light-speed cross-border transfer, and near-zero loss.

The most powerful insight is yet to come. Why is China uniquely positioned to execute this? The answer lies in two words. Electricity prices.

China is uniquely positioned to lead in the AI race because it has solved the “physical” constraint of intelligence: electricity prices. While algorithms are digital, running them requires massive amounts of power, and China’s ability to provide this power at a fraction of the cost in the West is becoming a decisive competitive edge.

Electricity for data centres in China can be as low as 3 cents per kilowatt-hour, roughly one-third the price in the U.S.. Unlike the U.S., where regional grids often operate with thin reserve margins, China maintains a deliberate surplus of electricity. This allows them to “soak up” the massive power demands of AI without destabilising the grid.

The State Grid Corporation of China plans to invest approximately 4 trillion RMB (US$579 billion) between 2026 and 2030 to further upgrade the power grid, specifically to support the future “intelligent economy”. Local governments often provide electricity subsidies for data centres, sometimes cutting power bills by up to 50% if they use domestic chips, further offsetting other costs.

In northwestern China, the situation is different. In specialized wind and solar power zones in Zhongwei, Ningxia, or Qingyang, Gansu, electricity prices can drop as low as 0.20 RMB (0.029 USD) per kilowatt-hour. This represents the absolute global price trough. The per-token cost gap between China and the U.S. can be seen from here.

Now you understand why DeepSeek API pricing can be nearly 20 to 30 times cheaper than OpenAI. This is not due to subsidies. This is not dumping. This is northwestern green electricity pushing cost advantages to their absolute limit within large language models.

Even more ingenious is the export mechanism for tokens. When you export electric vehicles, you face tariffs, trade barriers, and customs inspections at ports. Tokens travel via fiber optics. Under current WTO rules, electronic transmissions are temporarily exempt from tariffs. There are no containers, no cargo ships, and no customs declarations. Chinese electricity, cloaked in data, walks boldly into every terminal device worldwide. This is, without question, the strongest strategic backdoor available for China’s energy strategy.

Now consider another set of data that may surprise you. Recent statistics show that 4 out of the top 5 models on OpenRouter are Chinese large models, including MiniMax’s M2.5, Moonshot AI’s Kimi K2.5, Zhipu’s GLM-5, and DeepSeek’s V3.2. Their combined consumption reaches 85.7%. Chinese AI models have evolved from followers to price setters. This is only the beginning.

NVIDIA CEO Jensen Huang has long predicted that the inflection point for the AI Agent era has arrived. In the future, a single AI completing a task may consume 10 to 50 times as many tokens as it does today. Institutional forecasts project that by 2030, China’s AI inference token consumption will grow from 100 trillion in 2025 to 390,000 trillion by 2030. The ceiling for demand is not even visible yet.

So what is the essence of this transformation? Throughout human history, every reconstruction of the great-power order has begun with a revolution in the form of energy. The British Empire rose on coal and steam. The United States rose on oil and internal combustion. Today, China is quietly rewriting the rules through the ultimate coupling of electricity and computing.

Those northwestern green power resources that once had to be curtailed, causing heartache due to the inability to absorb them, are now being repriced and redeployed as tokens. Previously, we exchanged sweat for foreign exchange. Now, we exchange algorithms for foreign exchange. This is not overtaking on a curve. This is switching to an entirely new track.

Have you noticed? The changes that truly reshape the world often do not happen in headlines. They happen when an ordinary person opens a chat window on their phone, types a line of text, and waits for a reply. Behind that moment lies the wind of Inner Mongolia, the hydropower of Sichuan, and the sunlight of Xinjiang. They travel thousands of kilometers, burn inside GPUs, transform into tokens, cross the Pacific, and land on their screen.

What we are exporting is not merely data. It is the confidence of a civilization.

After reading this, do you believe token exports represent the smartest strategic move in China’s energy history? Pay attention. This is just the beginning.

Source: https://www.benzinga.com/Opinion/26/03/51533819/china-cant-export-electricity-so-it-did-something-smarter-the-ai-token-revolution-explained

Anndy Lian

www.anndy.com

j j j

March 19th, 2026

Clarity Without Complacency: Why the SEC-CFTC Framework Is a Start, Not a Finish Line

The March 2026 joint framework from the Securities and Exchange Commission and the Commodity Futures Trading Commission represents the most significant regulatory development in U.S. crypto history. While most of my peers see this as “good”, I view this moment with cautious optimism.

The classification of 16 major digital assets, including Bitcoin, Ethereum, Solana, and XRP, as digital commodities under primary CFTC jurisdiction finally provides the legal certainty that institutional capital has demanded.

Clarity, however welcome, does not equate to perfection. The framework’s very structure reveals tensions that could undermine its stated goal of fostering innovation while protecting investors.

Order Meets Oversight Gaps

The 5-category taxonomy, covering Digital Commodities, Digital Securities, Digital Collectibles, Digital Tools, and regulated Payment Stablecoins under the GENIUS Act, offers a pragmatic scaffold for a market that has operated in a regulatory gray zone for too long.

By acknowledging that assets can transition from securities to commodities as decentralization deepens, the agencies have embraced a dynamic view of technological evolution that the static Howey test never accommodated. This is progress.

The practical implications of shifting oversight from the SEC’s disclosure-heavy regime to the CFTC‘s market-conduct focus raise legitimate questions about investor safeguards.

Commodities regulation simply does not mandate the same level of financial transparency, audit requirements, or fiduciary obligations that securities law imposes.

For retail participants who have grown accustomed to the SEC’s investor-first posture, this represents a tangible reduction in recourse should manipulation or fraud occur. The data bears this out. While the CFTC has expanded its enforcement capabilities, its budget and staffing remain a fraction of the SEC’s, limiting its capacity to police a market now valued in the trillions.

The GENIUS Act’s Safeguards Could Backfire

The GENIUS Act’s treatment of stablecoins illustrates another layer of complexity. While the legislation rightly mandates one-to-one reserve backing, monthly attestations, and segregation of customer funds, it explicitly prohibits issuers from paying yield on stablecoin holdings.

This well-intentioned guardrail against shadow banking risks inadvertently pushes yield-seeking users toward unregulated offshore platforms or riskier DeFi protocols, potentially increasing systemic fragility rather than reducing it.

Furthermore, the Act’s bankruptcy provisions, while granting stablecoin holders super-priority status in theory, leave unresolved questions about the practical enforceability of those claims across fragmented custody arrangements.

If a major issuer were to fail, the FDIC’s $250,000 insurance limit applies to the corporate account holding reserves, not to individual token holders. This gap could leave millions of users exposed despite the framework’s consumer-protection rhetoric.

Perhaps the most pressing concern is the framework’s non-binding status. The SEC and CFTC do not legislate. Congress does. What we have today is an interpretive memorandum, not codified law, and as such, it remains vulnerable to shifts in agency leadership, judicial challenge, or superseding legislation like the pending Clarity Act.

Policy Without Law Leaves Investors Exposed

This uncertainty is compounded by the grey period inherent in the transition mechanism. Projects must now navigate costly legal analyses to determine precisely when they have achieved sufficient decentralization to shed their securities classification. For early-stage teams operating on lean budgets, this ambiguity could stifle the very innovation the framework purports to enable.

Moreover, national security experts at institutions like CSIS have warned that the GENIUS Act’s focus on centralized issuers may leave decentralized protocols and privacy-enhancing technologies outside the regulatory perimeter, creating vectors for sanctions evasion that adversaries could exploit.

From my vantage point, having engaged with both regulators and builders, I see this framework not as an endpoint but as a foundation on which more durable, adaptive regulation must be built. The harmonization of SEC and CFTC authority through Project Crypto is a historic step toward ending the jurisdictional turf wars that have long paralyzed U.S. crypto policy.

The Real Test Will Be in How Regulators Apply

Still, true regulatory maturity requires more than asset classification. It demands ongoing dialogue with technologists, economists, and civil society to ensure that rules evolve alongside the systems they govern. The inclusion of on-chain activities like staking, mining, and wrapping within the framework’s analytical scope is encouraging.

The devil will be in the implementation details that regulators now must develop through notice-and-comment rulemaking. The market has responded positively to the clarity, with institutional interest in the newly designated digital commodities rising measurably since the announcement. But we must resist the temptation to declare victory prematurely.

The framework’s success will ultimately be judged not by the elegance of its taxonomy but by its real-world outcomes. Does it reduce fraud without stifling experimentation? Does it protect consumers without cementing incumbent advantages?

Does it position the United States as a leader in responsible digital asset innovation, or merely as a jurisdiction that has replaced one set of uncertainties with another?

Prioritize Transparency and User Protection

As we await Congressional action to codify these principles into law, the industry must remain engaged, constructive, and vigilant. Builders should leverage the newfound clarity to prioritize transparency and user protection, not as a regulatory checkbox but as a competitive advantage.

Investors must recognize that commodity classification does not eliminate risk and should conduct due diligence accordingly. Policymakers must continue to listen to the diverse voices shaping this ecosystem, from developers in decentralized autonomous organizations to consumer advocates demanding accountability.

Do not get me wrong. The March 2026 framework is a big plus for the industry, yes, but it is a plus that comes with asterisks. It is a map, not the territory. It is a starting gun, not a finish line. Those of us who have championed decentralization, privacy, and financial inclusion for over a decade understand that regulatory clarity is necessary but insufficient.

Classification to Cultivation

The work now shifts from classification to cultivation. We must build the institutions, standards, and cultural norms that will allow digital assets to fulfill their promise without repeating the excesses of traditional finance.

If we approach this moment with both appreciation for the progress made and humility about the challenges ahead, the United States can yet lead the world into a more open, equitable, and innovative financial future. The framework gives us the rules of the road. It is up to all of us to ensure the journey delivers on its destination.

Source: https://www.financemagnates.com/cryptocurrency/clarity-without-complacency-why-the-sec-cftc-framework-is-a-start-not-a-finish-line/