Four sentences. Strip the adjectives, the framing, and the engagement bait, and that is the entire verifiable payload of a story that moved through crypto media this week: a proposed US bill, aimed at crypto investors who misreported in the past, creating a dedicated disclosure program, under which qualifying taxpayers receive reduced penalties.
That's it. No bill number. No sponsor. No committee referral. No retroactivity clause. No definition of "qualifying." No number attached to "reduced."
Which makes the most important sentence in the whole story the one almost everyone skipped: the bill reduces penalties. It does not waive them. Reduced. Not waived. Every headline I have seen this week quietly upgraded that word into "amnesty," and that single substitution is the difference between a tax bill and a tax fantasy.
So is this a genuine off-ramp for people who made honest mistakes, or a discount coupon printed to look like a pardon?
I have spent fourteen years reading contracts — smart ones, and now legislative ones — and the rule is identical in both environments. Code is law, but audits are the truth we chase. The only audit available here is a forensic reading of four clauses' worth of fog. So let's do it properly: what the bill actually does, what it structurally cannot do, and who ends up holding the bag once the headlines wear off.
Context: the enforcement clock is the story
Set the timeline first, because the timeline is why this bill exists at all.
America's crypto tax enforcement has been running on a ten-year delay, and 2025 is the year the delay expires.
Form 1099-DA — the broker reporting form built specifically for digital assets — enters service for transactions executed in 2025, filed in 2026. The first wave carries gross proceeds. Basis reporting phases in behind it, for assets acquired from 2026 onward. Separately, Rev. Proc. 2024-28 killed universal pooling: from tax years beginning in 2025, basis must be tracked wallet by wallet, address by address, with unused basis allocated across accounts by a deadline rather than netted globally.
Read those two developments together and you get something the industry has been studiously not saying out loud: the IRS is about to be handed a machine-readable ledger of gross proceeds for millions of retail accounts, while the taxpayer is simultaneously being forced onto a per-wallet accounting standard almost nobody was actually following. The Automated Underreporter program does one thing well — it matches third-party reported numbers against what you filed and mails a notice when they diverge. It does not care about your intent. It cares about arithmetic.
This is not hypothetical. In November 2017 a federal court granted the IRS a John Doe summons against Coinbase, initially seeking records on roughly 500,000 accounts, ultimately narrowed to about 14,000 accounts with transactions above $20,000 in a single year. That summons was the opening shot. The 2019 and 2020 rounds of "educational letters" to tens of thousands of holders were the warning. 1099-DA is the enforcement.
So the bill arrives at a specific moment, and the timing is the tell: it lands precisely when data-matching capability outruns the population's record-keeping ability. That is the political condition under which governments build pressure-release valves. Not charity. Capacity management.
There is a bear-market dimension too, and it is the part that actually costs people money. In a drawdown, holders who owe back taxes and penalties liquidate to pay. The liquidation is forced, it lands at the worst possible price, and it is triggered by an administrative liability rather than a market view. Any mechanism that reduces the size of that forced-sale overhang is a marginal improvement in market structure — which is exactly the size of the claim you should make about this bill. Marginal. Structural. Not a catalyst.
One more piece of scaffolding, because it matters for everything below. The US already has a voluntary disclosure mechanism — the IRS Criminal Investigation Voluntary Disclosure Practice, restructured in 2018. It works, roughly, like this: come forward before the Service contacts you, cooperate fully, file six years of returns, pay tax, interest, and a civil fraud penalty on the highest-liability year. In exchange, CI generally will not recommend prosecution. It is not an amnesty. It has never been an amnesty. Nobody calls it one, because the people who use it have lawyers and the people who don't have headlines.
The proposed bill is, in the most charitable reading, a legislative attempt to formalize and soften that mechanism for a specific population. In the least charitable reading, it is a press release with a legislative wrapper.
Compare the sequencing with Europe, because the contrast is instructive. The EU built reporting first — DAC7, then DAC8 for crypto-asset reporting, with the OECD's CARF framework landing behind it — and is only now getting around to arguing about penalties. Washington is doing it backwards: talking relief before the reporting infrastructure has produced its first mismatch notice. Between the hype cycle and the blockchain reality, the ordering of the two tells you which system is being built by engineers and which by press offices.
The word that carries the entire bill
Start with arithmetic, because arithmetic is where "reduced" stops being a modifier and becomes a number.
US penalties on an underpayment stack; they do not replace. The accuracy-related penalty under §6662 is 20% of the underpayment. Civil fraud under §6663 is 75%. Failure to file under §6651(a)(1) runs 5% of unpaid tax per month, capped at 25%. Failure to pay under §6651(a)(2) adds 0.5% per month, also capped. Estimated tax shortfalls carry their own charge under §6654. Interest accrues on top of all of it at the federal short-term rate plus three points for individuals — and interest is not a penalty. No bill, no executive order, no act of Congress waives interest.
Now run a plausible retail case. A trader realizes $400,000 of gains across 2021 and reports $120,000, because his exchange export double-counted an internal transfer and he never reconciled it. He pays tax on the lower figure. The underpayment is roughly $60,000 at federal rates before state tax. Add a 20% accuracy-related penalty — $12,000 — plus failure-to-pay accruals, plus three years of compounding interest.
If this bill reduces the penalty from 20% to 5%, the taxpayer saves $9,000 on a bill still well north of $75,000. If it reduces 75% fraud exposure to a 20% accuracy-related penalty — the more meaningful scenario, and the one the "honest mistake" framing points at — the savings are genuinely large, but only for someone who can survive a fraud investigation in the first place.
That is the whole ballgame, and the source reporting does not give us the number. Anyone telling you this is a "way out" without a percentage attached is selling a headline, not a calculation.
There is a second gap in the reporting that matters just as much, and it is categorical rather than numerical. The bill as described addresses past filing errors — people who filed and got it wrong. It says nothing about non-filers, who are a different population, a larger one, and the only one with meaningful criminal exposure. A disclosure program that covers mis-filers but not non-filers is a program for people whose paperwork was sloppy, not for people whose paperwork was absent. Those two groups need completely different legal representation and completely different relief. Conflating them is the single most common error in the coverage so far, and the source text does not resolve it.
Basis is a data problem before it is a legal problem
Here is where my day job intrudes, and where I think the coverage has been most naive.
In the summer of 2020 I audited a yield aggregator's interest calculation module before mainnet — the specific Solidity lines that compounded rewards per block. I found a logic flaw, contacted the team directly, and urged a delay that they ultimately took. What I remember most clearly is not the bug. It is that the module was computing a number that would eventually become somebody's taxable income, using no audit trail, no event emission granular enough to reconstruct it later, and no documented rounding convention.
Multiply that by every protocol in the market and you have the actual root cause of the "honest mistake" problem. The reason crypto filings are wrong is not primarily that people are evasive. It is that the underlying accounting was never designed to be reconstructed.
Wallet-by-wallet basis under Rev. Proc. 2024-28 turns that from an inconvenience into a forensic exercise. Consider a mid-sized holder in 2025: forty addresses, six chains, three bridges, a wrapped-asset position, a staking delegation, and a DeFi LP that rebalanced forty times. Under the new rules she cannot pool. She must be able to state, per address, what came in, at what basis, and what went out — and she must be able to prove a bridge transfer was a transfer rather than a disposition. That last point alone has sunk more filings than any other single issue I have encountered.
Then there is the layer-2 problem, which nobody in this debate is discussing. I have written for two years that "decentralized sequencing" is a PowerPoint — that the sequencer on most major rollups is, functionally, a single centralized node operated by a foundation. The tax consequence is concrete and ugly: when the sequencer is one node, the canonical history of your L2 activity lives on that foundation's indexer. If that indexer migrates, re-indexes after an upgrade, or retires an explorer, your basis on that chain becomes a research project. I have personally burned a weekend reconstructing a year of rollup transactions from raw calldata because an explorer's schema changed without notice.
There is a smaller, more insidious L2 issue layered on top: gas. Every time you pay a rollup fee in ETH, you have disposed of a fractional amount of ETH and recognized a fractional gain or loss. Thousands of micro-dispositions, per user, per year. Nobody reports them. The de minimis relief that would make this sane has been proposed in Congress repeatedly and has never passed. The speed of news is fast, but the chain is slower — and the chain is generating taxable events faster than any human filing process can absorb them.
The 2023 depeg nobody reported
Now the part I genuinely have not seen written anywhere, and the part that should worry you most about the accuracy of your own history.
Stablecoins are treated — in almost every retail mental model, and in most tax software defaults — as cash equivalents pegged at exactly $1.00. That assumption is wrong, and it has been wrong in ways that generated real, unreported, quantifiable taxable events.
On 11 March 2023, USDC traded down to roughly $0.87. Anyone who swapped USDC into USDT or ETH during that window recognized a loss on the USDC — a real loss, at a real basis, reportable. Anyone who swapped back at par after the peg restored recognized a gain. Both legs are dispositions. Almost nobody tracked them, because almost nobody held a basis record for a stablecoin in the first place.
And this is where the settlement layer of the entire market becomes a tax problem, not merely a reserve problem. USDT clears the majority of non-USD pairs in this industry. Tether's reserves have never been subjected to a genuinely independent, institutional-grade attestation of the kind the market pretends exists — the industry has collectively decided not to look, and has been deciding that for years. That is an old argument. Here is the new one: for tax purposes, the entire non-USD trading universe is priced through an unaudited constant. Every gain you have ever computed on a BTC/USDT pair assumes the denominator was worth precisely one dollar at the instant of the trade.
If that peg ever broke materially, the basis math for millions of trades would be retroactively wrong — and no disclosure program on earth can fix a denominator that lied.
Disclosure is a subpoena you volunteer for
The structural problem with every voluntary disclosure regime is not a bug. It is the design.
To qualify, you submit a complete history. Six years, typically. Account statements, exchange exports, on-chain addresses, and increasingly the wallet-ownership proofs that let an examiner tie an address to a person. You sign consents extending the statute of limitations — Form 872 and its variants — because no revenue authority grants relief on a window it can no longer audit. And you do all of it in a proceeding where the same submissions can later support a civil determination, an information return, or, at the margin, a criminal referral.
The civil-criminal boundary is not decorative. Willful evasion under 26 U.S.C. §7201 and false returns under §7206(1) are felonies. Foreign account non-reporting under the FBAR regime carries willful penalties that can reach half an account balance, plus a separate criminal track under 31 U.S.C. §5322. The word "willful" is doing enormous work in every one of those provisions, and "honest mistake" is its mirror image.
So the bill's central term is not a moral category. It is an evidentiary one. It means: can you demonstrate a good-faith attempt to comply? Which requires records. Which means the program selects, almost perfectly, for the people who least need it.
Program, not substance
One final precision, because it is the boundary that separates real reform from a procedural bandage.
The bill, as described, establishes a procedure. It creates a channel and adjusts a penalty. It does not define a single substantive tax rule. Every genuinely contested question in crypto taxation survives it untouched.
Whether staking rewards are income at receipt at fair market value — the position in Rev. Rul. 2023-14, which the Jarrett litigation challenged and then abandoned as moot after the Service refunded — is unchanged. Whether an airdrop is income at receipt under Rev. Rul. 2019-24 is unchanged. Whether an LP position is a series of dispositions on every rebalance is unchanged. Whether an NFT falls under the 28% collectibles rate of §408(m) is unchanged. Whether a DAO treasury distribution to a contributor is compensation, a gift, or a capital event — for a multisig with no legal entity, no withholding, and no information return — is unchanged and, in practice, unresolvable.
That last one deserves a hard look, because it is where the basis hole is deepest. Governance token compensation from an entity that does not legally exist produces no 1099, no withholding, and no issuer-side record. The contributor is told the token is worth whatever the chart said on the day it vested, by whoever is loudest in the reply thread. This is the same pathology as governance itself. Delegation was supposed to decentralize decision-making; instead it concentrated it in a small set of loud accounts, because most holders do not research and simply delegate to whoever performs confidence. Valuation by vibes, governance by vibes. The tax bill for vibes arrives later, with interest.
A procedural amnesty applied to a substantively undefined system does not produce clarity. It produces a population that is compliant in form and exposed in substance. Sifting through the wreckage of a bull market, that is the pattern I keep finding: the paperwork gets fixed first, and the definitions get fixed never.
The uncomfortable read
Now the part that cuts against the framing, and the reason I do not think this bill is really about taxpayers at all.
The bill's actual customer is the revenue authority, not the investor. Think about the failure mode if nothing happens. 1099-DA lands. Automated matching fires on millions of accounts. A meaningful fraction of those mismatches are not evasion — they are the basis-reconstruction failures described above, spread across people who genuinely tried. The administrative cost of adjudicating that population case by case is enormous. The political cost of mass penalty assessments on retail is worse. And the litigation risk is real, because a penalty assessed against someone with a good-faith but unreconstructable record is exactly the fact pattern that produces bad case law for the government.
A disclosure program converts that chaos into a queue. Pressure-release valves exist because somebody upstream did the math on what happens without one.
Which leads to the clause I expect to find when the text is actually published, and which the reporting does not mention: a limitations waiver. Every disclosure regime I have studied requires the taxpayer to extend the assessment window. It is the price of admission. If this bill carries one — and structurally it almost certainly must — then the relief is being purchased with an expansion of the government's reach into years that would otherwise be closed. That is not a scandal. It is standard. But it belongs in the headline, and it will not be.
Second point, and this one cuts against my own instinct to be generous: the failure scenario is worse than the bill's absence. The most dangerous outcome is not that the bill dies. It is that it lingers — introduced, referred, amended, folded into a larger tax package, reported again in six months, then again in twelve. A taxpayer who defers a current filing, or skips an existing voluntary disclosure, on the expectation of future relief is betting on legislative timing against a statutory deadline. The legislative calendar runs in years. Filing deadlines run in April. Anyone who treats "the bill is coming" as a reason to delay a current-period obligation is trading a certain cost for a speculative one. I have watched this exact behavior in protocol governance: holders defer a decision because a vote is "coming," and the deferral itself becomes the decision.
Third, the coverage is the story. This bill received four sentences of factual reporting. Compare that to the word count generated by a mid-tier token listing, or by any ETF headline. The information gain here is the inverse relationship between coverage depth and individual risk: the stories that can send you to prison get the least ink, because there is no chart attached.
What to watch
Three things, in order of diagnostic value.
The eligibility definition. Not the penalty percentage — the proof standard. Whether the bill contains an explicit safe harbor for taxpayers who can document a good-faith reconstruction attempt, or leaves "honest mistake" to case-by-case discretion, determines whether this is a real off-ramp or a discretionary lottery. Watch for the phrase "reasonable cause" and watch precisely what it is attached to.
Committee assignment and co-sponsorship. A tax-relief bill without bipartisan co-sponsorship in the House Ways and Means Committee or the Senate Finance Committee is a press release. Standalone penalty-relief legislation in the US has a miserable historical passage rate. The ones that survive get folded into reconciliation packages, where they are also most likely to be narrowed into something smaller than advertised.
The commercial tell. The real signal is not legislative at all. If a major exchange ships a historical tax reconstruction product in the next two quarters — bulk export, per-wallet basis allocation, bridge mapping, L2 indexing — then the industry is pricing this bill correctly, as a services event rather than a market event. That is where the money is. Tax tooling and on-chain forensics are the only unambiguous winners in this story, and they win whether or not the bill passes, because 1099-DA is already law.
For everyone else, the operational advice is boring and non-negotiable. Do not read "reduced" as "waived." Do not defer a current-period filing on the strength of a bill at introduction stage. Do not self-file into a program that may require you to extend the statute of limitations on years you would rather leave closed — get a practitioner to price the criminal exposure first. And start rebuilding your basis now, from exchange exports and raw chain data, because that work has to happen regardless of what Washington does.
The ledger doesn't forget. It has never forgotten a single one of your transactions — not the bridge, not the rollup fee, not the stablecoin you swapped at eighty-seven cents and told yourself was still a dollar.
The only open question is whether you can afford to make it testify.