MEV on Trial: When Crypto Trading Bots Break the Law

In a Manhattan federal courtroom in late 2025, twelve jurors spent three days arguing over a question no American court had ever been asked to settle: when a trader reorders transactions on a public blockchain to pull money out of other traders, is that clever arbitrage or is it wire fraud? They could not agree. The judge declared a mistrial, and the two defendants, brothers Anton and James Peraire-Bueno, walked out of court still facing a retrial that, by the schedule reported after a later hearing, cannot begin before November 2026.

The thing they were accused of stealing has a name most people outside crypto have never heard: MEV, short for Maximal Extractable Value. It is the profit a sophisticated actor can squeeze from the power to decide which transactions enter a block and in what order. In 2026 MEV is no longer a fringe curiosity. It has become a multi-billion-dollar industry with its own supply chain, its own oligopoly, and now its own criminal docket. Ethereum changes hands near $2,425 and Solana near $105 as this piece goes out, according to CoinGecko, and on both chains a quiet tax rides along with almost every swap.

MEV is not one thing, and that is the heart of the problem. Some of it keeps markets efficient and lending protocols solvent. Some of it skims a fraction of a percent off ordinary traders who never notice. A sliver of it looks, at least to prosecutors, like theft. This guide sorts the strategies from the benign to the toxic to the allegedly criminal, explains the machinery that makes them possible, and asks why the law is still struggling to draw a line that the code never did.

What MEV Is, and Why It Will Not Disappear

MEV is the value a block producer, or someone paying a block producer, can capture by adding, removing, or reordering transactions, over and above the standard block reward and gas fees. The idea was named in a 2019 paper, Flash Boys 2.0, whose authors warned that competition to capture this value could grow fierce enough to threaten the stability of the chain itself. The acronym first stood for miner extractable value; after Ethereum switched from mining to proof of stake in 2022, the M was redefined as maximal, because validators, not miners, now assemble blocks.

The reason MEV cannot simply be switched off is that it grows out of two features users actually want. The first is a transparent, public mempool, the waiting room where pending transactions sit in the open before they are confirmed. The second is a market that lets scarce block space flow to whoever values it most. Put those together and you get a race: anyone who can see your trade before it settles, and who can pay to be ordered ahead of or behind it, has a chance to profit. Ethereum’s own developer documentation walks through the simplest version, in which a token trades at two different prices on two exchanges and a bot buys low on one and sells high on the other inside a single transaction.

The scale is large and still growing. Cumulative MEV extracted on Ethereum runs well into the billions of dollars, and by several tallies sandwich attacks alone account for more than half of it. None of this is hypothetical lab work. It is a live, professionalised business that touches most active addresses whether they realise it or not, which is exactly why the question of whether any of it is illegal has stopped being academic.

A Spectrum, Not a Monolith

The public conversation treats MEV as a single villain, but practitioners do not. Inside the industry the strategies fall along a spectrum that runs from value that makes markets work better, through value that is merely redistributed from one party to another, to value that is taken from an identifiable victim who ends up clearly worse off. That spectrum, more than any single definition, is what the courts, the regulators, and the protocol designers are all circling in 2026.

The table below maps the main strategies onto that spectrum. The right-hand column is the one that has shifted most this year: a strategy’s legal exposure now depends less on how much it earns and more on whether a prosecutor can point to a deceived victim and a rule that was broken.

The Benign End: Arbitrage and Liquidations

Start at the healthy end. Arbitrage is the oldest and least controversial form of MEV. When the same asset trades at two prices, a bot that buys the cheap one and sells the dear one pockets the difference and, in the process, drags the two prices back together. On a decentralised exchange that is not a parasite on the system; it is the mechanism that keeps a pool’s price honest. The same logic governs stablecoin arbitrage, where bots defend the one-dollar peg of a dollar token by trading whenever it drifts. As stablecoins become the settlement layer of choice across more jurisdictions, including the new Asian regimes covered in our guide to 2026 stablecoin rules, that peg-defending arbitrage only grows more important.

Liquidations are the second benign pillar. Lending protocols require borrowers to stay overcollateralised. When a position slips below its threshold, anyone can repay part of the debt and seize the collateral at a discount. The bot that wins the race earns the bonus, but the service it performs, clearing bad debt before it threatens the protocol’s solvency, is one the system is built to reward. Without fast liquidators, a sharp price drop could leave a lending pool holding unbacked loans while depositors scramble to withdraw.

Both are MEV in the strict sense, since both depend on seeing and ordering transactions. Yet neither has a victim in the ordinary meaning of the word. The arbitrageur profits from a gap that should not exist; the liquidator profits from enforcing rules the borrower agreed to. This is the part of the spectrum almost nobody wants to regulate, and for good reason: remove it and the markets on top of it stop working.

The Gray Zone: Backruns, JIT Liquidity and the Long Tail

Move one step along and the moral clarity fades. Backrunning places a transaction immediately after a known pending one, usually to capture the arbitrage that the first trade creates. A large swap pushes a pool out of line; the backrunner trades right behind it to restore balance and bank the difference. In most cases the original trader is no worse off than they would otherwise have been, which is why backrunning sits close to the benign end. It is also the engine behind many of the user-friendly tools described later, which deliberately auction the right to backrun a trade and hand most of the proceeds back to the person who made it.

Just-in-time (JIT) liquidity is murkier. Here a bot sees a large swap heading for a concentrated-liquidity pool, deposits a big slug of liquidity a moment before the trade, collects most of the trading fee, then withdraws immediately afterwards. The swapper often gets a slightly better price, so it is not a simple mugging. The losers are the passive liquidity providers who kept their capital in the pool through quiet markets and volatile ones alike, only to watch a bot parachute in, skim the fattest fees, and leave. Whether that counts as predatory or merely as efficient competition is a question the industry still has not settled.

The long tail covers everything opportunistic: sniping a token the instant it lists, racing to claim an airdrop, grabbing an underpriced NFT the moment it surfaces, front-running a freshly deployed contract. Some of this is harmless speed. Some of it shades into outright deception, especially where bots exploit naive contracts or bait a victim into a losing transaction. The long tail is where MEV most often blurs into what a lawyer would simply call a scam, which is why it is worth keeping separate from the pure ordering games that give the field its name.

The Toxic End: Sandwich Attacks and the Retail Tax

At the toxic end sits the sandwich attack, the strategy that earns MEV its bad name and supplies the clearest victim. The recipe is simple. A bot spots your pending swap in the public mempool. It buys the same asset just before you, pushing the price up; your order then fills at that worse price; and the bot sells immediately after, pocketing the spread. You are the filling. Most sandwiches shave somewhere between a few tenths of a percent and about one percent off a trade, small enough that the victim rarely notices and large enough that, repeated across tens of thousands of swaps a day, it compounds into real money.

The most notorious practitioner is a bot known as jaredfromsubway.eth, which has run sandwiches at industrial scale for years. In the spring of 2026 it even sandwiched Vitalik Buterin, wrapping a roughly $4 token swap by Ethereum’s co-founder inside about $1.14 million of its own volume (and, as it happened, barely moving the needle on that particular trade). The same bot became a cautionary tale weeks later when an attacker drained roughly $7.5 million from it using a swarm of fake token contracts, proof that the predators are themselves hunted.

Sandwiching is where the victim cannot be argued away, which makes it the strategy most exposed to the market-manipulation theories discussed below. It helps to be precise about what kind of threat it is. A sandwich does not steal your keys or drain your wallet; it is an ordering attack, not a custody attack, a different animal from the physical-coercion and seed-phrase risks we covered in our hardware wallet reviews. No amount of cold storage protects you from a sandwich. Only transaction privacy does.

CEX-DEX Arbitrage: The Darkest Corner

The single most lucrative form of MEV is also the hardest to see. CEX-DEX arbitrage exploits the price gap between a centralised exchange, where most price discovery still happens, and an on-chain pool that lags it by a block or two. When Bitcoin or Ethereum moves on a big centralised venue, the on-chain price has not caught up yet, and a searcher who trades the DEX against a position held on the exchange captures the difference. Because one leg of the trade sits off-chain, the strategy is called non-atomic: it cannot be packed into a single, all-or-nothing transaction, so it carries real inventory risk, and the profit never lands neatly on-chain for anyone to tally.

A detailed 2025 academic study, Measuring CEX-DEX Extracted Value and Searcher Profitability, put numbers on that shadow. Across nineteen months it counted more than $233 million captured from over 7.2 million CEX-DEX arbitrages, run by only nineteen searchers, with the top three taking roughly three-quarters of the value. For much of the period this non-atomic flow outweighed all the classic on-chain MEV combined. The authors called it the darkest corner of the dark forest, and the label fits: this is a capital-intensive game dominated by a handful of professional trading firms, invisible to the retail user whose pool price it quietly leans on.

CEX-DEX arbitrage sits awkwardly on the spectrum. It does not sandwich anyone, and over time it helps on-chain prices track the wider market. Yet it also concentrates MEV in the hands of the few players who can run low-latency infrastructure across both worlds, and it is the main reason the business has hardened into something that looks far more like high-frequency trading than open competition.

The Case That Put MEV in a Courtroom

Every strategy above operates within the rules of the game, however unfair some of them feel. The Peraire-Bueno case is different, and that is exactly why it matters. In April 2023 the two brothers allegedly did something no ordinary searcher does: they broke the trust assumption at the heart of Ethereum’s block-building pipeline.

The mechanics were intricate, but the outline is clear enough. Ethereum validators that outsource block construction rely on software called MEV-Boost and on intermediaries called relays, which are meant to keep a proposed block’s contents secret until the validator has committed to publishing it. Prosecutors say the brothers planted bait transactions to lure sandwich bots, then exploited a flaw in how the relay released block data to see a block’s full contents ahead of time, rearrange them, and siphon about $25 million out of those bots in roughly twelve seconds. The Department of Justice indicted them in May 2024 in what it billed as the first criminal case built around MEV.

The charges are the familiar federal trio for financial wrongdoing: conspiracy to commit wire fraud, wire fraud, and conspiracy to commit money laundering, brought by the US Attorney’s Office for the Southern District of New York. Notice what is absent: there is no securities-fraud count, no claim that a token was an unregistered security. The government’s theory is simpler and, for crypto, more unsettling. It argues the brothers deceived the validators and the protocol by posing as honest participants while planning all along to break the rules, a theory the defense mocks as an invented honest validator standard that exists nowhere in the code.

That framing forces a question the industry usually ducks: when does exploiting how a system works become a crime? A smart-contract hack that drains a protocol through a coding bug is one thing, and security firms exist to catch those; the audit business has even grown into a trust layer of its own, as our profile of CertiK explored. But the Peraire-Bueno exploit did not break a smart contract. It broke a social and software convention about how blocks ought to be built, and no audit report covers that.

The first trial, in late 2025, ended in deadlock. After three days of deliberation the jury could not agree, and the judge declared a mistrial. Prosecutors signalled they would retry the case; the defense asked the court to dismiss the charges outright, arguing the government never proved a crime. However the retrial lands, the first attempt showed how hard it is to get twelve ordinary people to agree that reordering public data is theft.

The Eisenberg Precedent and the Code-Is-Law Problem

The defense’s sharpest weapon is a ghost from an earlier case. In October 2022 a trader named Avraham Eisenberg drained roughly $110 million from Mango Markets, a Solana-based lending platform, by using the protocol’s own rules to inflate the value of his collateral and borrow against it. A jury convicted him in April 2024 of commodities fraud, commodities manipulation, and wire fraud. Then, in May 2025, US District Judge Arun Subramanian vacated those convictions, granting a motion for acquittal on the ground that prosecutors had not shown Eisenberg made any materially false statement. In a permissionless system, the judge reasoned, doing what the code allows is not by itself a lie. Prosecutors have appealed, and that appeal was still live in 2026.

That ruling is the shadow over the Peraire-Bueno retrial, and the brothers’ lawyers have leaned on it hard. At a hearing this year, James Peraire-Bueno’s attorney Patrick Looby told the court that there was nothing in the lure transactions themselves that made a promise about the relative order of transactions, and argued that, as with Eisenberg, neither Ethereum nor MEV-Boost carried any terms of service his clients could have breached. The parallel is hard to miss: if using a permissionless system as designed is not fraud, how can exploiting its quirks be?

The case has pulled in the policy world. The advocacy group Coin Center filed an amicus brief objecting to the honest validator theory, arguing that honest validation in these networks is a mathematical check rather than a moral or legal one, and that the brothers contravened no clear protocol rule. Peter Van Valkenburgh of Coin Center called the prosecution a massive overstep and warned that a conviction would, in the group’s words, massively chill public participation in these systems. The DeFi Education Fund filed a letter in the same spirit. Not everyone is convinced; plenty of observers think that reordering other people’s trades to grab $25 million in seconds is exactly the kind of conduct fraud law should reach. That split is why the first jury hung.

Where the SEC and CFTC Draw the Line

The criminal cases are only one front. The deeper question is whether MEV, or at least its predatory end, runs afoul of the civil market-abuse rules the SEC and the CFTC already enforce. The two obvious hooks are the SEC’s Rule 10b-5, the catch-all against fraud and manipulation in securities markets, and the CFTC’s Rule 180.1, its near-identical anti-manipulation rule for commodities. A 2022 legal analysis, Blockchain Transaction Ordering as Market Manipulation, argued that some MEV, sandwiching above all, could plausibly meet the test for manipulation if a court is willing to treat the abuse of privileged control over trade ordering as the on-chain cousin of front-running a client.

The trouble is jurisdiction and reach. The SEC polices securities; the CFTC polices commodity derivatives and spot-market manipulation; most tokens being sandwiched are not cleanly either, and the two agencies have spent years arguing over who owns the territory. Even where a rule plainly applies, enforcement hits a wall: the sandwich bot is an anonymous address on a permissionless exchange, with no license to revoke, no office to subpoena, and often no identifiable human behind it. You can write a rule against a ghost, but you cannot easily serve it.

The SEC’s own posture has shifted, too. After years of what critics called regulation by lawsuit, the agency has leaned toward writing rules instead of litigating every edge case, the change we traced in our report on how the rulebook is replacing the lawsuit. That makes a sudden MEV crackdown unlikely, even as the wider Washington agenda, mapped in our regulatory countdown, keeps market-structure questions firmly on the table. One influential submission to the SEC’s crypto task force, from the firm Paradigm, argues that the base layer where ordering happens should be treated as neutral infrastructure rather than a venue the securities laws should supervise directly. For now the clearest legal pressure on MEV comes not from a regulator’s rulebook but from a criminal courtroom, and only for the single strategy that broke the pipeline itself.

The Assembly Line: How a Block Gets Built

To understand why one strategy ended in court while the rest did not, it helps to see the machine that turns MEV into money. Since the Merge, most Ethereum validators no longer build their own blocks. A four-step supply chain does the work instead. Searchers run the bots that spot opportunities and package transactions into bundles. Builders assemble those bundles, along with ordinary transactions, into the most profitable block they can. Relays sit in the middle as trusted couriers, passing the finished block to the validator while keeping its contents hidden until the last moment. And the proposer, the validator whose slot it is, simply signs the most valuable block on offer.

The software that stitches this together is MEV-Boost, built by the research collective Flashbots, and more than nine in ten Ethereum blocks now flow through it. The design was meant to democratise MEV, letting any validator, even a hobbyist running a node at home, auction block space to professional builders and keep a share of the proceeds. It worked. It also created the exact trust assumption the Peraire-Bueno brothers allegedly exploited: the relay has to behave honestly, because for a brief moment it holds information worth millions.

The market that grew on top of MEV-Boost is anything but decentralised. A snapshot of block-building market share shows how few hands hold the pipe.

A single builder, Titan, produces more than half of all Ethereum blocks, and the top three together build close to nine in ten, according to relayscan.io. The relay layer is only slightly less concentrated, split mostly between the ultrasound and Titan relays. This matters beyond fairness. Because builders decide what goes into a block, they can also decide what stays out: after US sanctions hit the Tornado Cash mixer in 2022, a large share of blocks began excluding sanctioned transactions, a trend still tracked publicly at MEV Watch. Concentrated block building is concentrated censorship power, which is why Flashbots has pushed a shared, multi-operator builder called BuilderNet that returns value to users; it now accounts for about 15% of blocks.

If the Law Will Not Protect You, Code Might

If regulators cannot easily reach anonymous bots, users are left to defend themselves, and in 2026 they finally have a real toolkit. It comes in three layers.

The first layer is privacy. The simplest defense against a sandwich is to keep your transaction out of the public mempool altogether. Private order-flow services and protected RPC endpoints send your trade straight to a builder without broadcasting it for bots to see first. If the predators cannot see your swap, they cannot wrap it. Most major wallets now ship a private or protected routing option, and switching it on is the single most effective thing an ordinary trader can do.

The second layer is redistribution. Order-flow auctions flip the model: instead of letting a bot take the backrun for free, they auction the right to it and refund most of the proceeds to the trader. Flashbots’ MEV-Share and the widely used MEV Blocker both work this way, handing back the bulk of any value your trade creates. CoW Protocol goes further still, settling trades in batch auctions at a single uniform price, so that within a batch there is no ordering left to exploit and a sandwich becomes structurally impossible.

The third layer moves the fight into the application itself. Newer designs let a protocol capture its own MEV and hand it back to the people who generate it. Application-specific sequencers such as Sorella Labs’ Angstrom, and priority-ordering schemes on chains like Unichain, reclaim value that would otherwise leak to searchers and route it to liquidity providers or back into the protocol. The principle is the same across all three layers: if MEV cannot be abolished, at least make sure the person it is taken from gets most of it back.

The Protocol Fix: ePBS, FOCIL and Encrypted Mempools

The defenses above are patches on a system whose plumbing still depends on trusted relays. The longer-term answer is to rebuild the plumbing so no trusted middleman is needed, and that work is well underway. Ethereum’s next major upgrade, nicknamed Glamsterdam, carries a change called enshrined proposer-builder separation, or ePBS, specified in EIP-7732. It bakes the builder-to-proposer handoff directly into the protocol and removes the off-chain relay as a point of trust, closing the very category of exploit the Peraire-Bueno case turned on. Glamsterdam reached the Sepolia test network on 6 October 2026, with a mainnet launch expected later in the year but not yet dated.

ePBS alone is not a cure, and the people building it say so. In March 2026 Vitalik Buterin laid out a broader plan to curb builder centralisation, warning that separation by itself still leaves too much power in the hands of a few builders. His prescription pairs ePBS with inclusion lists, a scheme called FOCIL specified in EIP-7805 that forces builders to carry transactions they might otherwise censor, plus encrypted mempools and anonymised routing so bots cannot read a trade until it is too late to attack it. Encrypted mempools in particular would strike at sandwiching directly: a bot cannot front-run a transaction it cannot decrypt.

None of this arrives overnight, and some of it may slip. But the direction is set. The protocol is slowly absorbing the defensive tricks that users and applications have been bolting on from the outside, which is the healthiest possible outcome: fewer trusted intermediaries, less to exploit, and a smaller surface for the next criminal case to form around.

MEV Beyond Ethereum: Solana and the Layer 2s

MEV is not an Ethereum problem; it is a feature of any blockchain where ordering has value, and the second-largest venue handles it very differently. Solana has no global public mempool, which removes the shared waiting room where Ethereum’s sandwich bots hunt. Instead, most of its MEV flows through Jito, whose client software runs on the large majority of Solana’s stake and lets searchers submit bundles with tips to win favourable ordering. The JitoSOL liquid-staking token recycles some of those tips back to stakers, turning a chunk of MEV into yield rather than a pure leak.

Solana has also been readier to police the toxic end directly. In 2024 the Solana Foundation removed dozens of validators from its delegation program for running private mempools that enabled sandwich attacks on retail users, and a foundation official, Tim Garcia, described the purge as final. Where a network has a coordinating body willing to act, social enforcement can do what securities law struggles to.

Ethereum’s layer-2 rollups are a third model again. Most run a single sequencer that decides ordering, which kills the public-mempool free-for-all but concentrates ordering power in one operator who can, in principle, extract MEV themselves. Some are experimenting with auctioning the right to order transactions; others with shared sequencers spread across many operators. The common thread across Ethereum, Solana, and the rollups is that MEV never disappears. It only migrates to wherever ordering power is concentrated, which is precisely where the next legal and design fights will be.

The Bottom Line

Strip away the courtroom drama and the roadmap acronyms, and MEV in 2026 reduces to a single idea: whoever controls the order of transactions controls a stream of value, and that stream will be captured by someone. The useful question is never whether MEV exists but who gets it, and at whose expense. Arbitrageurs and liquidators earn it by making markets work. Sandwich bots take it from people who did nothing wrong. And a thin band of conduct, the kind that breaks the pipeline itself, has now crossed from aggressive trading into alleged crime.

For the ordinary user, the practical takeaway is more encouraging than the headlines suggest. The toxic end of MEV is the part you can most easily opt out of: route your trades privately, use a batch-auction venue, and the sandwich tax largely vanishes. For the industry, the harder work is structural, and it is happening, as trusted relays give way to enshrined separation and encrypted order flow. The law will keep circling the edges, the Peraire-Bueno retrial will plant an early marker, and the design will keep shrinking the surface the law has to reach. MEV is not going away. But in 2026, for the first time, the people it preys on have real choices, and the people who abuse it have reason to watch the courthouse.

Frequently Asked Questions

Is MEV illegal?

Most MEV is legal. Arbitrage, liquidations, and backrunning operate within the rules of the networks they run on, and no US regulator has declared them unlawful. The open question concerns the predatory end, such as sandwich attacks and exploits that break the block-building pipeline. The first criminal MEV prosecution, against the Peraire-Bueno brothers, ended in a mistrial in late 2025 and is headed for a retrial, so the law here is genuinely unsettled.

What is a sandwich attack and how do I avoid it?

A sandwich attack is when a bot sees your pending swap, buys the same asset just before you to push the price up, lets your trade fill at that worse price, then sells immediately after for a profit. The most reliable defense is to keep your transaction out of the public mempool by using a wallet with private or protected routing, or by trading through a service such as MEV Blocker or a batch-auction venue such as CoW Protocol, where sandwiches are structurally blocked.

How much does MEV actually cost ordinary traders?

On a single swap, a sandwich usually costs between a few tenths of a percent and about one percent of the trade, so most people never notice it. Across the whole market the totals are large: cumulative MEV extracted on Ethereum runs into the billions of dollars, and sandwich attacks make up more than half of it by some measures. It is best understood as a small, recurring tax on active on-chain trading rather than a one-time loss.

Will Ethereum upgrades like ePBS end MEV?

No single upgrade ends MEV, because MEV exists wherever transaction ordering has value. Enshrined proposer-builder separation, arriving with the Glamsterdam upgrade, removes the trusted relay and closes the specific exploit behind the Peraire-Bueno case. Pairing it with inclusion lists, encrypted mempools, and anonymised routing would blunt sandwiching further. The goal is to shrink and redistribute MEV, not to abolish it.

What was the Peraire-Bueno case about?

Anton and James Peraire-Bueno are brothers accused of exploiting a flaw in Ethereum’s MEV-Boost software in April 2023 to view and reorder the contents of a block before it was finalised, draining about $25 million from sandwich bots in roughly twelve seconds. Prosecutors charged them with wire fraud and money laundering conspiracy. The defense argues they merely used the system as built, echoing the overturned conviction of Mango Markets exploiter Avraham Eisenberg. The first trial ended in a mistrial.

By Yuki Tanaka, senior DeFi correspondent at HOGE Wire.