Who Controls AI Agents? The Future of Ownership and Currency Sovereignty | WebX2026

WebX2026 | Session Report

Who Controls AI Agents? Detailed Report on the WebX2026 Panel

Yat Siu × Sunayna Tuteja × Toshiyuki Morita

On July 13, at the Binance stage of WebX2026, a panel discussion titled "Who Controls AI Agents? - The Intersection of IP, Regulation, and Innovation" was held. The panel featured Yat Siu from Animoca Brands, Sunayna Tuteja, former Chief Innovation Officer of the Federal Reserve, and Toshiyuki Morita, Head of APAC at Perplexity AI. The discussion covered the actual use of AI agents, ownership, policy responses, and the impact of dollar-denominated stablecoins on currency sovereignty, all within a 37-minute session.

Panelist Profiles

Yat Siu
Animoca Brands
Co-founder & Chairman
A technology entrepreneur and investor based in Hong Kong. He co-founded Animoca Brands in 2014 and has invested in over 570 companies in the Web3 and blockchain space. Since 2018, he has been known as an early advocate for blockchain and NFTs.

Sunayna Tuteja
Federal Reserve System
Former Chief Innovation Officer
As the first Chief Innovation Officer of the Federal Reserve, she led the introduction of technologies such as AI, digital assets, and tokenization, along with their policy impacts. She previously established the digital asset division at TD Ameritrade.

Toshiyuki Morita
Perplexity
Head of APAC
He oversees the APAC region for Perplexity AI, responsible for the rollout strategy of consumer AI products, including the AI search and agent-based browser "Comet."

Alex Shin
Moderator
Founder/CEO of Shinlabs
An entrepreneur and investor who has advised major Web3 foundations such as Solana, AVAX, Sui, Uniswap, and MakerDAO. He co-founded the blockchain VC "Hashed" and established Shinlabs in 2022 after serving as CBO.

The Year of AI Agents: Consumer and Enterprise Use

Moderator Alex Shin opened the discussion by stating, "2026 is the year AI agents became a reality. They shop, browse, negotiate, and sometimes even get restrained by courts," and turned to Morita to discuss consumer usage.
Toshiyuki Morita
At Perplexity, we initially operated with a very Silicon Valley mindset, prioritizing consumers as a small organization. We built trust based on accuracy without hallucinations, which allowed us to transition from regular chat to the first AI agent-based browser, "Comet." We are now moving into the realm of "computing," where agents handle tasks on behalf of users.
Alex Shin pressed, "Is token consumption really leading to productivity improvements?" He referenced reports that companies implementing workforce reductions and agents have not seen the expected results.
Toshiyuki Morita
I believe there are still significant applications, but we see that there are even greater potentials that have yet to surface. The cost per task should dramatically decrease.
Yat Siu
Token prices are generally trending downward. In our on-chain products, switching from Gemini to MiniMax's M3 resulted in a cost reduction to one-twentieth for equivalent performance. What used to cost $50 to $100 a month is now down to $1 to a few dollars. Of course, deep research requires options like Grok or Gemini Pro, but for everyday tasks like calendar management and email responses, the Chinese frontier models are more than sufficient. In Europe, there is still strong caution towards Chinese models, but in reality, they can operate on private clouds with open weights. The trend of AI is moving towards the edge, where delicate processes like email handling will run local LLMs on home machines, while deep research will be handed off to Claude or Perplexity.
As usage becomes mainstream, consumption will spike. If one spends $1,000 or $2,000 a month, it’s not a problem as long as more value is generated. In fact, many people still do not understand orchestration and how to utilize it effectively. The "harnesses" and sub-agents provided by Claude, Gemini, and Perplexity are the current IP, and companies will increasingly build their own harnesses and prompts to hold as their assets.

What Policymakers Need: "Hands-On Experience"

Alex Shin asked Tuteja whether such private sector movements could also apply to the government sector.
Sunayna Tuteja
Before joining the Fed, I spent most of my career in a position that was somewhat adversarial to the Fed (from the perspective of the crypto and fintech industry). While it might seem that central bank innovation has a completely different playbook, it is actually quite similar; the only difference is the pace. The attitude I recommend to policymakers is "responsible innovation." The approach of "breaking things and apologizing later" should not and cannot apply to policymakers. However, being indifferent is also not acceptable. One cannot wait ten years to see how things unfold. It is essential to get into the field and gain hands-on experience.
ChatGPT emerged in November 2022, and by January of the following year, all 20,000 Fed employees were enabled to use AI (then ChatGPT). Even Hugging Face was activated in a safe isolated environment that does not touch core systems like Fedwire. It was emphasized that not only developers but also staff from various operational lines such as payments and research should understand the technology. In the world of cryptocurrencies, I have often raised eyebrows at guidance from policymakers who have never used a MetaMask wallet. If one has not used the technology, they cannot create coherent and practical policies.

The Sovereignty Debate Over Models - Mistral and "Mythos/Fable"

Alex Shin asked Yat Siu and Toshiyuki Morita, who have interacted with policymakers in various countries, about their impressions of the understanding on the policy side.
Toshiyuki Morita
There are differences between countries, but U.S. policymakers are quite positive. Many countries are moving towards sovereign AI models, and they are much more proactive than I had imagined. Even Japan's new strategy on AI and copyright for 2026 seeks to explore ways to better utilize AI without overly restricting the government, businesses, or the general public, while still valuing existing frameworks.
Yat Siu
The level of understanding at the policy level is clearly higher for AI than for cryptocurrencies and digital assets. This is because AI is more familiar and everyone is actually using it. However, a theme that has particularly emerged this year is national security and data sovereignty. This is notably seen in parts of Asia and Europe. Japan may find it hard to grasp this as it does not have its own domestic LLM. However, when talking to government officials in Europe, such as in Austria or Germany, I have repeatedly encountered situations where they almost force the use of Mistral simply because it is "European-made." When selling AI services to the government, whether to use a U.S.-made model, a Chinese one, or a European one—considerations that did not exist a few years ago are now routinely questioned as part of digital sovereignty and data sovereignty. Will Mistral become the "Airbus" for Europe? Will the Chinese model become a barrier for China? In the future, there is a possibility that governments will impose restrictions on which models they can use. I believe the recent incident surrounding "Mythos/Fable" has also had a significant impact on these discussions. However, this topic had already been under the surface before that. The question is whether it is truly safe for a country to continue relying on external models. While Japan's policies are unclear, many other countries are just now beginning to discuss this issue.
Alex Shin said, "I would like to hear some candid opinions," and turned to Sunayna Tuteja.
Sunayna Tuteja
This is an unprecedented area, and everyone is moving forward while learning by trial and error. I just came from Washington D.C., where I was precisely in the midst of such discussions. Much of the current decision-making is based on rapidly changing data and shifting motives and incentives. I like Charlie Munger's saying, "Show me the incentives, and I will predict the actions and outcomes," and it is indeed crucial to focus on those incentives. I heard the same concerns when I was in Europe two weeks ago. There were similar voices during the cloud era, but they were smaller and did not gain significant traction. Regarding AI, those voices are much louder and the perspectives are broader. The industry also needs to continue its proactive engagement with regulators and policymakers. The biggest lesson from the world of cryptocurrencies, taking the discussions in the U.S. as an example, is that we did not start dialogues with regulators and policymakers quickly enough. This has solidified a narrative unfavorable to cryptocurrencies, and there are still messages that cannot be dispelled. The builders in this room also need to persistently engage with policy-making venues, even when the doors are hard to open. My ideal is for engineers and policymakers to sit side by side.

Stablecoins × AI Agent Economy ------ "MetaMask for Machines"

Alex Shin pointed out Japan's stablecoin deregulation and South Korea's cautious stance, asking Yat Siu how policymakers view the combination of AI and stablecoins, seeing blockchain as the next UX frontier for the AI agent economy.
Yat Siu
A few days ago, U.S. Treasury Secretary Janet Yellen mentioned the future of agent-based commerce, and it is clear that the U.S. is steering towards agent-based commerce and trying to lead in that field. In this context, blockchain and stablecoins become key. Current blockchain technology already allows for frictionless, almost zero-cost transfers of dollar-denominated value in a decentralized manner, handling transactions on a massive scale. However, it remains difficult for end users to use. They still have to remember 12 or 24-word seed phrases—this is a challenge for the industry. However, this is not a problem for AI agents. In fact, the software, which can best leverage the inherent strengths of blockchain as programmable money, is the agent itself. Programmers have made money as "translators" between humans and machines, but soon machines will be able to write code themselves. Everyone in this room is already a programmer, and we have entered an era where agents can take over that work.
If the mechanism of programmable money expands to a scale of trillions of dollars, it will effectively provide every AI agent with a bank account and payment method. Blockchain will become the ideal "banking system for machines" for AI agents. Considering that in the next five years, each person will have 3 to 5 AI agents, it is expected that in the future, hundreds of billions of agents will engage in economic activities. To truly make them autonomous, they need to have money. It may be a while before mega banks provide bank accounts to AI agents, so instead, they will likely have cryptocurrency wallets. Just as people in the Philippines who could not have bank accounts became "bankable" through cryptocurrencies, machines will also be integrated into the economy through cryptocurrencies. When hundreds of billions of agents each have wallets and can even shop for daily necessities, will it be Visa, IBAN transfers, or a reliable and scalable unique payment and settlement layer?—we believe that it will all be realized on top of cryptocurrencies.
This will change the very way we see the world. If you use Perplexity or Gemini, you might wonder when was the last time you clicked through to a website. Two years ago, you would have gone to the site every time, but now you don’t. You are satisfied with the answer itself. Perplexity is one of the few conscientious services that properly indicates links (sources), but still, very few people actually check the sources. This means that traffic is significantly declining for all companies that have websites. The world that relied on advertising revenue—where players like Google, Facebook, and Meta earned through click prices as microtransactions—is transitioning to an economy where everything is traded through scaled microtransactions. We believe that the foundation for realizing this is blockchain.

Who owns it, and who is responsible?

Alex Shin asked Mr. Morita whether there is a movement to actually entrust agents with the collaboration with banks, based on discussions about blockchain and MetaMask.

Mr. Morita
One of the reasons we started with agent-based commerce is precisely related to payments. The ongoing debate is about the timing of "when to approve" and "how much authority to grant." When fans want to buy, when do we allow the agent to make the purchase? When do we conduct bank collaboration and authentication? This has been an unresolved issue for a long time. Last week, we announced an MOU with SMBC, Nippon Steel, and the SBI Group, and began efforts to incorporate our API into a new trading platform. What we consistently think is that the decision-making process should always be placed on the user side, not on AI. Rights and regulatory responsibilities should remain with the user. Last week, I heard a story about someone being automatically enrolled in a higher plan with an LLM service, resulting in a high charge on their credit card, and I want to avoid such situations. It is crucial to place all decision-making processes on the user side.

In response to Mr. Morita's point about "user sovereignty," Alex Shin shifted the discussion to ownership and repeatedly questioned Mr. Yat about the nature of ownership and management when having multiple agents.

Mr. Yat Siu
How much discretion to grant is no different from how much discretion you give to the people working under you or your children. Some may remember when the App Store first came out, and children clicked on in-app purchases in Farmville, leading to high charges. I describe AI agents as "six-year-old PhD holders." They possess incredible intelligence, but at their core, they are still children, and common sense needs to be provided by us. Therefore, I have no resistance to giving my agent $100 or $50 on-chain to let them operate freely. Whether it succeeds or fails, it is part of a fun experiment. On the other hand, I do not want to allow them to trade $100,000 without permission. In the case of "HelloMinds," Superior Trade has built a harness, allowing agents to trade on Polymarket and Hyperliquid, while retaining final authority on the human side and ensuring the custody of funds and the safety of transactions. When it comes to ownership, what matters is not the LLM itself, but your workflow—your "harness." Whether it is part of edge computing or how models are combined, it becomes the IP that companies and individuals should own. When thinking about AI agents, what people have achieved through OpenAI is just the beginning. We are currently at a stage where we are following the same evolution as when we set up our own web servers in the 90s, which eventually led to outsourcing to Amazon.

Alex Shin sought Mr. Tuteja's perspective from a regulatory standpoint regarding ownership when agents take action or create something.

Ms. Sunayna Tuteja
I am very bullish on agent-based commerce, and I want to add one more point in addition to the reasons everyone has mentioned. If agent-based commerce works, the foundation of its payments will be dollar-denominated stablecoins. That is why it is called the "GENIUS Act." During my time at the FRB, I began to pay attention to this theme as the next big challenge. We need to consider three junctions: tech stack, policy stack, and trust stack. While discussions on the technical side are progressing, the policy stack also needs to be redesigned, and the same goes for the trust stack. The framework for fraud detection in today's payments does not apply in the world of agent-based commerce. In simple terms, if a machine tries to conduct a transaction now, it is automatically detected as fraud. However, the world of agent-based commerce is one where machines transact with humans or machines transact with each other. Redesigning digital IDs will also be necessary. The market structure itself will need to be reviewed as the speed and volume of payments will be drastically different. These challenges are not just homework for policymakers; they are also a treasure trove for builders, and new problems that next-generation startups need to solve lie here.

How Dollar-Denominated Stablecoins Will Redefine Currency Sovereignty
As time was running out, Alex Shin prefaced his question by saying, "This will be a tough question," and asked what dollar-denominated stablecoins mean for other national currencies.

Mr. Yat Siu
I will say this with the caveat that it is my personal opinion, but I believe the aim of the dollar-denominated stablecoin strategy is to further strengthen the hegemony of the dollar in the world. Over 300 million people in Southeast Asia, Africa, and South America are already using dollar-denominated stablecoins in their daily lives, and they are a generation that has grown up with the dollar. Countries may emerge that no longer need local currencies, following paths like Argentina. If billions of people primarily use dollar-denominated assets, the U.S. will further solidify its position to effectively tax the entire world through quantitative easing. Europe is particularly vulnerable to this trend. It is already in a state of being "colonized" by U.S. tech companies, and the same concerns are being raised in the AI field. Regarding stablecoin policy, Europe is still focused on discussions around CBDCs (central bank digital currencies) and does not have its own effective stablecoin policy. There should be nothing more sovereign than one's own currency. Like sovereign AI, each country needs to have its own stablecoin. CBDCs do not work because no one trusts them. South Korea is an exceptional case in that its capital movement is relatively closed, similar to China and Saudi Arabia, but open markets like Japan are strongly affected by this flow. In the future, countries will advance their responses from the perspective of licensing systems, and agent-based commerce will bridge that gap. Stablecoins from Hong Kong, stablecoins from the U.S., stablecoins from Japan—once the rules and regulations are in place, AI agents will seamlessly process transactions behind the scenes, so users won’t even need to be aware of it.

Ms. Sunayna Tuteja
To answer briefly as time is limited, I am a self-proclaimed "stablecoin advocate," and I take pride in having been involved in the GENIUS Act during my time at the FRB. Ultimately, the customer—the end user—holds the final decision-making power. If customers want dollar-denominated stablecoins, that is what will be chosen. There is no stopping that. However, many unanswered questions remain about how the spread of dollar-denominated stablecoins will affect other countries in the long term. This is a massive experiment even for U.S. monetary policy. I jokingly answered, but it is a serious topic that should be considered and questioned continuously, and I would like to continue this discussion with anyone in this room.

Session Summary
Mr. Morita summarized the current phase where discussions about AI and cryptocurrencies are converging.

"We are always in a flow that converges towards AI. The themes of AI and cryptocurrencies are also trying to converge. I believe the discussions around stablecoins this time represent a turning point where policy formation and stablecoins intersect."

From the actual use of AI agents to ownership, policy responses, and how dollar-denominated stablecoins shake currency sovereignty—the 37-minute panel left the question of "who holds the power" unanswered, leaving homework for the next discussion.

Related Articles
Event WebX 2026: Complete Guide to Dates, Tickets, and Highlights
Stablecoins: What are Stablecoins? A Comprehensive Explanation of Mechanisms, Types, Risks, and Future Potential
Exchanges: Recommended Comparison of Domestic Cryptocurrency Exchanges
![Follow on Google News](https://public.chaincatcher.info/upload/news/202607/d2d4ab0dd8cf487f893074503b1f11c4.png