Binance Agent OS lets ChatGPT, Claude, and other AI agents place trades across spot, margin, and futures through a single protocol. Five competitors launched similar systems in the past 30 days. The custody models are different, the liability language is almost identical, and nobody has answered the question that matters most: what happens when an agent loses money.

The largest cryptocurrency exchange in the world announced on Wednesday that AI agents can now trade on its platform. Not through a workaround, not through an unofficial API wrapper, but through a purpose-built system called Binance Agent OS that connects directly to the exchange’s markets, wallets, and execution engine.

The system uses Model Context Protocol, an open standard created by Anthropic that gives compatible AI applications a uniform way to plug into external tools. Binance listed Claude, ChatGPT, Codex, and VS Code among the agents that can connect. Once linked and granted permission, an agent can pull live market data, check balances, and place trades across spot, margin, convert, and futures products.

Binance is not the first exchange to do this. It is the fifth major platform to launch agent-trading infrastructure in less than 30 days. But it is the largest, and the architecture it chose reveals something about where the industry thinks risk actually lives.

What Binance Agent OS actually does

Agent OS bundles four components that previously required separate integrations into a single access layer. The first is the exchange’s existing API, which handles market data and order execution. The second is an agent-focused wallet hub that creates and manages isolated sub-accounts. The third is x402, a payment protocol layer that handles fee routing and micropayments between agents and services. The fourth is a skills marketplace where developers can publish and discover pre-built trading strategies that agents can load and execute.

At the center of the system sits a new Binance MCP Server. MCP is an open standard that lets AI applications connect to external tools without users juggling API keys locally. An agent running on a user’s machine or in the cloud connects to the MCP Server, requests access to specific capabilities, and operates within the scope the user grants.

The skills marketplace is the component that distinguishes Agent OS from a simple API upgrade. Binance had already shipped seven AI Agent Skills in March 2026, covering spot trading, USD-margined futures, margin trading, Alpha market data, wallet data, execution tools, and asset management. Agent OS wraps these skills into a discovery layer where any compatible agent can browse, evaluate, and activate strategies without the developer writing custom integration code.

This means a user does not need to program a trading strategy. They can point an AI agent at the skills marketplace, describe what they want (“rebalance my portfolio to 60% Bitcoin, 30% Ethereum, 10% stablecoins every Monday”) and the agent selects and executes the appropriate skills. The gap between intention and execution has collapsed to a single sentence.

The critical design choice is the sub-account architecture. Every agent operates through what Binance calls an “Agentic sub-account,” a walled-off partition of the user’s holdings. The sub-account can receive funds from the main account but cannot send them anywhere external. If the agent is compromised, stolen, or simply makes bad decisions, the damage is theoretically contained to whatever the user deposited into the sub-account.

Binance also chose not to grant agents withdrawal scope. An agent can buy, sell, convert, and open leveraged positions, but it cannot move assets to an external wallet. This is the single most important guardrail in the system, and it is worth understanding exactly what it does and does not protect against.

What it protects against: an agent draining funds to a third-party address. What it does not protect against: an agent making a series of bad trades that reduce the sub-account balance to zero, or opening leveraged positions that get liquidated. The guardrail prevents theft. It does not prevent loss.

The five competitors and their custody models

Binance is not building in isolation. Five other platforms launched agent-trading products between July and August 2026, and each made fundamentally different choices about where risk sits.

Coinbase rolled out a tool in late July that lets agents trade and make payments. Coinbase is also funding agent-focused startups through its Base accelerator program, signaling a long-term commitment to the category. The custody model mirrors Binance: exchange-hosted, with agent access scoped to specific capabilities. But Coinbase went further by integrating agents directly into its Base Layer 2 network, creating a path for agents to interact with on-chain protocols without leaving the Coinbase ecosystem. A Coinbase-connected agent can, for example, provide liquidity to a decentralized exchange on Base, claim yield, and reinvest the proceeds, all without the user touching a wallet.

Gemini introduced its own agentic trading feature in June. Gemini’s approach is the most conservative of the group. Agent access is restricted to read operations and spot trading only, with no margin or futures capability. The exchange positions this as a safety-first approach, arguing that agents should prove reliability on simple tasks before gaining access to leveraged products. Critics counter that the restrictions limit the utility enough to make agents impractical for anything beyond simple rebalancing, which is precisely the type of task that did not need an AI agent in the first place.

MetaMask took the opposite approach by launching a self-custodial AI wallet. In this model, the agent holds its own private keys and operates autonomously on-chain. The user sets spending limits and asset restrictions, but the agent can interact with any decentralized protocol within those bounds. This is the highest-risk, highest-flexibility option. If the agent’s key management is compromised, there is no exchange to freeze the account. The funds are gone in the same way they are gone when any private key is stolen: irreversibly.

MoonPay built agent products specifically for Telegram, targeting the messaging platform’s large crypto-native user base. MoonPay agents can execute purchases, check balances, and manage portfolios through conversational commands. The custody model is MoonPay-hosted, similar to the exchange models but with a payment processor’s compliance infrastructure underneath. The Telegram integration is significant because it meets users in a platform they already use daily, removing the friction of downloading a separate application or navigating an exchange interface.

Ledger and MoonPay jointly developed a system that lets users cap how much an agent can spend from a hardware wallet. This is the most novel approach in the group. The hardware wallet acts as a spending limit enforcer: the user approves a maximum transaction amount and a time window, and the agent can operate freely within those constraints. Once the cap is hit, the agent stops until the user physically approves a new allocation on the device. The elegance of the design is that the security guarantee comes from hardware, not software. Even a fully compromised agent cannot spend more than the user authorized on the physical device.

The range of architectures reveals an industry that has not converged on a standard. Exchange-hosted sub-accounts, self-custodial wallets, hardware-enforced spending caps, and payment-processor models are all live simultaneously, each making different tradeoffs between convenience, security, and user control.

The liability gap nobody is talking about

Every platform in the current wave shares one characteristic: the terms of service place the entire risk of agent-driven trading on the user.

Binance’s announcement included a disclaimer stating that use of its AI services is “at the user’s own risk” and that outputs “should not be relied on alone for decisions.” Binance also cautioned users to review each order and transfer before confirming, placing the responsibility for keeping an agent in check on the user rather than the exchange.

This language is standard across the industry. Coinbase, Gemini, MetaMask, and MoonPay all use variations of the same framework: the platform provides the infrastructure, the user assumes the risk, and the agent exists in a legal gray zone where it is treated as a tool rather than a fiduciary.

The problem is that agent trading is designed to be autonomous. The entire value proposition is that the agent acts without constant human oversight. Telling users to “review each order before confirming” while simultaneously building a system optimized for hands-off execution creates a contradiction that no platform has resolved.

Consider a scenario: a user connects an AI agent to Binance Agent OS, deposits $10,000 into the agentic sub-account, and sets the agent to execute a momentum-following strategy on Bitcoin futures with 10x leverage. The agent opens a long position at $77,000. Bitcoin drops 10% overnight. The position is liquidated. The $10,000 is gone.

Who is responsible? Under the current terms of service, the user is. The agent is a tool. Binance provided the infrastructure. The user chose the strategy, the leverage, and the allocation. But the user also chose to use an AI agent specifically because they did not want to monitor every trade manually. The terms of service and the product design are pulling in opposite directions.

Now consider a more complex scenario: the same agent, running the same strategy, opens a position that triggers a cascading liquidation across multiple accounts. The agent’s trade was the marginal order that pushed a thinly traded futures market past a liquidation level, forcing other positions to close, which pushed the price further, which triggered more liquidations. The user lost $10,000. Other traders collectively lost $500,000. The agent was following its instructions exactly as written.

In traditional finance, this type of cascade has clear accountability. The exchange’s risk management system should have circuit breakers. The broker should have position limits. The algorithmic trading firm should have kill switches. In crypto agent trading, none of these safeguards are required.

This is not a hypothetical concern. A U.S. survey published on Aug. 12 by BadCredit.org found that 79% of prediction market users lost money in the past year, with 51% using borrowed funds. Prediction markets and agent-driven trading are different products, but they share a common dynamic: automated or semi-automated decision-making systems that attract retail users who may not fully understand the risk surface.

Model Context Protocol and why it matters

The technical foundation of Binance Agent OS is Model Context Protocol, and understanding MCP is essential to understanding why this moment is different from previous waves of algorithmic trading.

MCP is an open standard created by Anthropic that gives AI applications a uniform interface for connecting to external tools. Before MCP, integrating an AI agent with an exchange required custom API wrappers, authentication flows, and error handling for each platform. A developer building a trading agent needed separate integrations for Binance, Coinbase, and every other exchange.

MCP changes this by creating a single protocol that any compatible agent can use to discover and interact with any compatible service. A Binance MCP Server advertises its capabilities (read market data, place orders, check balances) in a standardized format. An agent discovers these capabilities, requests access, and begins operating.

The implication is that agent trading will scale much faster than previous waves of automation. Building a trading bot in 2020 required weeks of API integration work. Building an agent-trading system in 2026 requires connecting to an MCP Server and writing a prompt. The barrier to entry has dropped by an order of magnitude.

This is both the promise and the risk. Lower barriers mean more participants, more liquidity, and more competition among strategies. They also mean more untested strategies, more inexperienced operators, and a higher probability of correlated failures when many agents react to the same market signal simultaneously.

The speed of adoption is already visible. Binance shipped its first seven AI Agent Skills in March 2026. Five months later, it launched a full platform with a skills marketplace, a sub-account system, and an MCP Server. The iteration speed suggests that agent trading is not an experiment for Binance. It is a core product strategy.

The flash crash question

The crypto market has a history of flash crashes driven by algorithmic trading. The May 2021 crash saw Bitcoin drop 30% in hours as leveraged positions were liquidated in a cascade. The FTX collapse in November 2022 triggered a similar dynamic, with automated selling amplifying human panic.

Agent trading introduces a new variable: agents that share underlying models. If a significant fraction of trading agents use the same foundation model (GPT-4, Claude, or their successors), they may develop similar market views and execute similar trades. This is not the same as traditional algorithmic trading, where each firm writes its own strategy. AI agents using the same model may converge on the same analysis and act in the same direction at the same time.

No exchange has published research on this correlation risk. No regulator has proposed rules for it. The closest precedent is the concern about passive index funds creating systemic risk by all holding the same stocks. But index funds rebalance on fixed schedules. AI agents can act in milliseconds.

Capital.com plans to introduce spot cryptocurrency services in the United Arab Emirates after its affiliate, Capital Vault, received a virtual asset license from the UAE Capital Market Authority.

Capital.com announced the approval on Aug. 21. The license authorizes Capital Vault to deal in virtual assets as an agent or matching principal and provide custody services for clients.

Once the service launches, eligible UAE customers will be able to buy and hold actual crypto assets through the Capital.com application. Capital Vault will handle trade execution, custody and settlement.

Capital.com did not disclose a launch date, list of supported cryptocurrencies, trading fees or minimum account requirements. The availability of every product may also depend on customer eligibility and local regulatory conditions.

Capital.com will offer ownership beyond CFD exposure

Capital.com currently provides crypto market exposure through contracts for difference in supported jurisdictions. A CFD tracks the price of an asset without transferring ownership of the underlying cryptocurrency to the trader.

Six weeks of bearish positioning ended in 24 hours. Here is how the trade unwound, who got caught, and whether the squeeze has legs.

Crypto derivatives markets had been building toward this moment for six weeks. Open interest in Bitcoin perpetual futures climbed steadily through July, with funding rates drifting negative as traders added to short positions. Bearish bets outnumbered bullish ones on every major exchange. On Binance, shorts held 51.64% of open interest. On OKX, 51.13%. On Bybit, 52.25%. The consensus was clear: the market was going lower.

Then, over the span of 24 hours, it went violently higher.

What triggered the squeeze

The first catalyst landed on Aug. 19 at approximately 2:30 PM UTC, when the U.S. Treasury announced it would at least double the maximum size of its liquidity support buyback operations for 10 to 20 year and 20 to 30 year nominal coupon securities. The cap moved from $2 billion to $4 billion per operation, effective Sep. 9 through Nov. 4.

Treasury buybacks are not quantitative easing. The department buys back illiquid, off the run bonds and replaces them with fresh, on the run issuance. The net effect on the government balance sheet is roughly neutral. But the market impact is not. By removing duration from the market, buybacks compress long end yields and improve liquidity conditions across risk assets. For more on the mechanics, see our breakdown of how the $4 billion Treasury buyback moved Bitcoin 8% in a day.

Bitcoin responded within minutes. The price moved from $64,100 to $66,800 in the first hour after the announcement. That initial move was enough to trigger the first wave of margin calls on leveraged shorts, and the cascade began.

The liquidation cascade

The mechanics of a short squeeze in crypto derivatives are straightforward but brutal. When a short position on a perpetual futures contract falls below its maintenance margin, the exchange liquidates it by placing a market buy order. That buy order pushes the price higher, which triggers more liquidations, which generates more buy orders. The feedback loop continues until the selling pressure from remaining shorts can absorb the forced buying.

On Aug. 19 and 20, the loop ran for roughly 18 hours before stabilizing.

Total liquidations across all major exchanges exceeded $3 billion. Short positions accounted for approximately $2.77 billion, or 92% of the total. Long liquidations were a rounding error at $264 million. According to CoinGlass data, roughly $1.29 billion in short positions closed within a single hour, the fastest concentrated squeeze of 2026. As we reported when Bitcoin first broke past $68K on the initial $1 billion short squeeze wave, the cascade was just beginning.

The breakdown by exchange reveals how concentrated the pain was. Binance saw approximately $518 million in liquidations. Hyperliquid, the decentralized perpetuals exchange that has grown rapidly this year, absorbed roughly $513 million. Bybit recorded around $303 million. The remaining liquidations spread across OKX, dYdX, and smaller venues.

Bitcoin shorts accounted for approximately $1.37 billion of the total, while Ethereum shorts contributed roughly $1.01 billion. The remainder came from altcoin positions, with Solana, XRP, and Dogecoin among the most affected.

The exchange level data reveals a secondary pattern that the headline numbers obscure. On Hyperliquid, a decentralized exchange that does not use a traditional order book for liquidations, the insurance fund absorbed roughly $47 million in losses during the cascade. The fund, which stood at approximately $380 million before the event, dropped to $333 million by the time the squeeze stabilized. On Binance, the auto deleveraging system activated twice during the peak liquidation hour, forcing profitable long traders to partially close their positions to cover the counterparty shortfall. These mechanisms prevented cascading failures at the exchange level but added to the speed and violence of the price move.

The altcoin liquidation data adds granularity that the Bitcoin and Ethereum headlines miss. Solana perpetual futures saw approximately $187 million in short liquidations, driven by the same macro catalysts plus the additional momentum from cumulative SOL ETF inflows crossing $1.16 billion earlier in the week. XRP shorts lost roughly $142 million, with the asset rallying 10% alongside the broader market. Dogecoin, which had seen a buildup of speculative short positions during a quiet July, contributed approximately $89 million. These figures matter because altcoin liquidations tend to be more violent per dollar of open interest. Altcoin perpetual markets are thinner, with fewer market makers and wider spreads. When liquidations cascade through these markets, the price impact per dollar liquidated is significantly larger than in Bitcoin or Ethereum.

Why the positioning was so extreme

The bearish lean in crypto derivatives markets did not appear overnight. It built over six weeks, from early July through mid August, during a period when multiple headwinds converged.

The CLARITY Act, the most comprehensive crypto market structure bill to reach the Senate floor, stalled after its procedural vote was postponed to September. The SEC finalized its “Regulation Crypto Assets” framework, which some market participants interpreted as an attempt to preempt Congressional legislation. For more on how these two frameworks conflict, see our analysis of SEC regulation crypto assets vs the CLARITY Act. Bitcoin had traded in a narrowing range between $60,000 and $66,000 since late June, with each rally attempt meeting selling pressure near the upper bound.

Funding rates on Bitcoin perpetual futures turned negative in late July and stayed negative through mid August, meaning that short traders were being paid to hold their positions. That dynamic attracted more shorts, creating a self reinforcing cycle of bearish positioning.

The numbers tell the story precisely. On Aug. 18, one day before the squeeze, the eight hour funding rate on Binance Bitcoin perpetual futures stood at negative 0.012%, a level that had persisted for three consecutive weeks. At negative funding, traders holding short positions receive a payment from traders holding long positions every eight hours. The payment is small in absolute terms but compounds meaningfully over weeks. A trader with a $10 million short position at negative 0.012% funding received approximately $3,600 per day simply for maintaining the position. That dynamic attracted capital into shorts not because of a directional thesis but because of the yield. When the forced unwind came, many of these yield seeking shorts had no thesis to defend and no plan for a stop loss.

The result was a market that was heavily one sided. When the Treasury announcement provided a fundamental reason for risk assets to rally, the positioning was too extreme to absorb the move without forced buying.

The second catalyst: the White House summit

The Treasury announcement alone might not have been sufficient to produce a $3 billion liquidation event. But it was followed within hours by reports that President Trump would host a crypto industry summit at the White House, attended by senior SEC officials and executives from major exchanges.

The summit, confirmed for late August, signaled that the administration remained committed to a regulatory framework favorable to the crypto industry. Coming on top of the Treasury buyback expansion, it created a second wave of short covering that pushed Bitcoin from $68,000 to above $71,000 on Aug. 20.

The combined effect of both catalysts was greater than either alone. The Treasury announcement provided the fundamental case for higher prices. The White House summit provided the narrative. Together, they forced the most aggressive unwind of bearish positioning since the collapse of FTX sent the market into a tailspin in November 2022.

How Ethereum outperformed

Ethereum’s 18% single day move was the standout of the squeeze. While Bitcoin gained roughly 8%, Ethereum outperformed by a factor of more than two. The reason lies in the composition of the short positions that were liquidated.

Ethereum shorts on major exchanges had grown disproportionately through July and August, partly because of skepticism about the Pectra upgrade timeline and partly because of persistent outflows from Ethereum spot ETFs. The net short positioning in Ethereum perpetual futures was, relative to open interest, more extreme than in Bitcoin.

When the squeeze began, Ethereum’s thinner order books amplified the price impact. Trading volume on Ethereum pairs surged 402% in 24 hours, according to AMBCrypto data. The asset moved from approximately $1,920 to above $2,270 before stabilizing near $2,250. For our full Ethereum price prediction, see our dedicated analysis.

The rally also exposed a structural risk in DeFi. On Aave, the largest decentralized lending protocol, just 9% of positions carry roughly half of the platform’s total debt. These positions are built around a leveraged Ethereum staking correlation trade, using WETH debt against liquid staking collateral like weETH, rsETH, and wstETH. The average health factor on these positions sits near 1.06, meaning an 8% to 9% wrapper discount could trigger a liquidation cascade on chain.

The staking correlation trade that dominates Aave’s risk profile operates on a simple premise that conceals significant complexity. A trader deposits weETH, a liquid restaking token issued by EtherFi, as collateral on Aave. The trader then borrows WETH against that collateral at a loan to value ratio near 90%. The borrowed WETH is restaked through EtherFi to produce more weETH, which is deposited again as collateral. Each loop multiplies both the staking yield and the leverage. At 10 times leverage, the effective annual yield on the trader’s equity approaches 40% to 50% before accounting for borrowing costs and gas fees. The trade is profitable as long as weETH maintains its peg to ETH within a narrow band. The moment the wrapper discount exceeds the health factor buffer, the entire recursive structure unwinds through liquidation.

The Aug. 20 rally did not trigger that cascade because ETH moved higher, not lower. But the concentration of risk in a small number of highly leveraged positions remains a vulnerability. If Ethereum corrects sharply from current levels, the same positions that survived the upside squeeze could face liquidation on the way down.

The institutional side of the trade added another layer to Ethereum’s outperformance. U.S. spot Ethereum ETFs, which had recorded net outflows for much of July and early August, posted net inflows of approximately $189 million on Aug. 19 alone. The reversal in ETF flows suggests that institutional investors were not only covering short positions in derivatives but also adding long exposure through regulated products. Weekly ETF inflow figures strengthened in tandem, signaling that the squeeze may have catalyzed a broader reassessment of Ethereum’s near term prospects among allocators who had been underweight the asset.

What the data says about follow through

Not every short squeeze leads to a sustained rally. The question is whether the forced buying created genuine demand or simply cleared out weak hands.

The evidence is mixed. On one hand, Bitcoin’s move above $72,000 broke a six week trading range and set a new short term high. Open interest has declined by approximately 15% since the squeeze, suggesting that leveraged positioning has been significantly reduced. Funding rates have turned positive, indicating that the market is no longer paying traders to be short.

On the other hand, the fundamental catalyst has a built in expiration date. The Treasury’s expanded buyback program runs only through Nov. 4, 2026. After that window closes, Treasury will reassess whether to maintain the larger operation size. If long end yields have stabilized by then, there is no guarantee that the program continues at its current scale.

Most Americans believe President Donald Trump and his family should not earn money from cryptocurrency while he holds office, according to a Reuters/Ipsos poll released on Aug. 19.

The survey found that 63% of respondents considered the Trump family’s crypto profits inappropriate. Another 32% viewed the activity as appropriate, while the remaining respondents did not answer the question.

Reuters and Ipsos conducted the nationwide online survey between Aug. 14 and Aug. 17. The poll included 1,166 U.S. adults and carried a margin of error of about three percentage points.

Trump crypto profits expose a partisan divide

Views differed sharply by political affiliation. About 69% of Republicans considered the family’s cryptocurrency earnings appropriate, according to the reported results. By comparison, 92% of Democrats said the activity was inappropriate.

The poll also examined broader concerns about private commercial interests. Around 69% of respondents said they believed Trump’s business interests influenced his presidential decisions. That included approximately two-thirds of independent respondents and nine in ten Democrats.

The findings measure public opinion and do not establish that Trump violated any law or influenced government policy for financial gain. The White House has consistently rejected allegations of a conflict.

“There are no conflicts of interest. The President only acts in the best interests of the American public,” White House spokesperson Anna Kelly told Reuters.

Trump has also said his investments are managed independently and that he does not participate in the family businesses’ daily operations.

Financial filing puts crypto income above $1.4 billion

The poll followed the publication of Trump’s annual financial disclosure in June. A Reuters analysis of the filing calculated that Trump reported more than $1.4 billion in income connected to cryptocurrency ventures during 2025.

The figure represents reported income rather than the current value of Trump’s personal cryptocurrency holdings. As crypto.news reported, the disclosure included more than $1 billion in crypto-related income from projects including World Liberty Financial and the Official Trump memecoin.

Companies linked to the Trump family received almost $800 million from World Liberty Financial activities, Reuters calculated. The total included more than $520 million connected to token sales and over $250 million from the sale of business interests.

The filing also reported approximately $635 million from licensing arrangements associated with the TRUMP token. In related coverage, blockchain analysis found that many buyers recorded substantial losses while Trump-linked entities continued receiving transaction-related revenue.

These figures should not be treated as a calculation of personal net profit. The disclosed revenue flowed through several companies and agreements, and some proceeds were shared among Trump family members and business partners.

Ethics concerns overlap with U.S. crypto policy

The debate comes as the Trump administration promotes legislation establishing clearer federal rules for digital assets. Lawmakers have disagreed over whether crypto market legislation should include restrictions on elected officials and their families.

As previously reported, proposed ethics provisions have become a central obstacle to advancing crypto legislation. Supporters argue that broad market rules remain necessary, while critics want stronger safeguards covering officials with financial interests in digital-asset businesses.

World Liberty Financial also received conditional approval on Aug. 14 to establish World Liberty Trust Company as a national trust bank. The Office of the Comptroller of the Currency listed the decision in its official records. Conditional approval does not allow immediate operations because the company must satisfy the regulator’s requirements before opening.

Congressional scrutiny, future financial disclosures and the conditions attached to World Liberty’s proposed trust bank will provide further tests of the separation between Trump’s public duties and family business interests. The Reuters/Ipsos results indicate that most Americans remain unconvinced that the current arrangements adequately address those concerns.

South Korea’s Shinhan Asset Management signed a four-party memorandum on Aug. 21 to test a Korean won tokenized fund using the Solana blockchain.

The agreement brings together Shinhan Asset Management, the Solana Foundation, tokenization platform Etherfuse and decentralized exchange Orca. The participants will conduct a proof of concept covering the fund’s issuance and distribution process.

The planned product would invest in short term Korean won bonds and target overseas institutional investors. However, the participants have not announced the fund’s size, expected yield or public launch date.

Shinhan will test the complete tokenized fund process

The proof of concept will examine the steps needed to issue and distribute a regulated tokenized fund. These include know your customer checks, anti-money laundering controls, token issuance and onchain liquidity arrangements.

Shinhan will provide asset management and regulatory expertise. Etherfuse will supply infrastructure for creating and managing the tokenized assets. Orca will help design the liquidity system used to distribute or exchange the fund tokens on Solana.

The Solana Foundation said the model draws from BlackRock’s BUIDL fund, one of the largest tokenized money market products. The comparison refers to the blockchain based distribution model. It does not mean the proposed Shinhan fund will hold the same assets or offer the same legal rights.

BlackRock’s product primarily invests in U.S. Treasury bills, cash and repurchase agreements. By comparison, Shinhan’s test concerns a Korean won product backed by short term domestic bonds.

Solana expands its institutional fund activity in Asia

Solana offers low transaction fees and rapid settlement, which could support frequent subscriptions, redemptions and transfers. However, those technical features do not remove securities registration, custody or investor protection requirements.

The network has already attracted other Asian asset managers. As previously reported, SBI Global Asset Management launched a tokenized Japanese equity fund on Solana in July. That product targets institutional and accredited investors through regulated tokenization platform DigiFT.

Shinhan has also been testing more than one blockchain. On Aug. 14, the asset manager signed a separate agreement with Plume to develop a demonstration for a won-denominated tokenized fund.

The parallel projects suggest Shinhan is examining different technical and distribution models rather than committing exclusively to Solana. Results from the tests could determine which infrastructure the manager uses after South Korea’s regulatory framework becomes effective.

Korea’s 2027 framework will determine any launch

South Korea’s National Assembly passed amendments supporting tokenized securities on Jan. 15. The legislation recognizes distributed ledgers as valid securities registries and permits qualifying investment contract securities to circulate through licensed securities companies.

The Financial Services Commission said issuers will still need to satisfy existing securities registration and disclosure obligations. Unlicensed companies will not be allowed to broker tokenized securities.

The amendments are expected to take effect one year after their promulgation, with implementation anticipated in early 2027. Regulators are preparing account management infrastructure and investor protection rules before the rollout, according to the FSC’s official statement.

As crypto.news reported, South Korean authorities are preparing rules for stocks, bonds and funds alongside a blockchain platform operated for the Korea Securities Depository.

Market forecasts remain uncertain

Solana said the existing tokenized real world asset market was worth about $36 billion. The announcement also cited a Boston Consulting Group projection suggesting the sector could reach as much as $30 trillion by 2030.

That projection should be treated as a forecast rather than an expected outcome. BCG’s more recent middle scenario estimates tokenized real world assets could reach $14 trillion by 2030 and $55 trillion by 2035. Its faster growth scenario places the market as high as $88 trillion by 2035.

Current adoption remains much smaller. In related coverage, publicly visible onchain assets grew to approximately $30 billion by mid-2026, led by private credit and tokenized government debt.

The next stage will involve completing the proof of concept and aligning the fund structure with Korea’s final rules. Any commercial launch will depend on regulatory approval, operational testing and demand from eligible overseas institutions.