A penalty is not a price until someone puts a number on it. Last week the Internal Revenue Service announced a discount and declined to name the figure.
Jarod Koopman, who runs IRS Criminal Investigation, told the UCLA Extension Tax Controversy Conference that the agency intends to rebuild its voluntary disclosure machinery: cases closed in 120 days instead of a year or more, penalties set "significantly lower," installment agreements and offers in compromise made routine, launch targeted before December. Buried in the same remarks was the line that matters more than all the others โ digital assets will not receive a separate disclosure track. Crypto taxpayers will queue with everyone else.
That is the whole story. Not leniency. Throughput.
The voluntary disclosure practice is the narrow door through which a non-compliant taxpayer walks before the government finds him. Under the framework the IRS set out in its November 2018 memo, a successful disclosure means a pre-clearance request, six years of amended or delinquent returns, full payment of tax and interest, and a civil fraud penalty assessed at 75% of the highest single year's unpaid balance. In exchange, criminal exposure evaporates โ provided the disclosure is truthful, timely, and complete. The old Offshore Voluntary Disclosure Program closed in September 2018 and was replaced by exactly this: a standardised, penalty-heavy, permanent channel.
For a decade that channel was tolerable because the population flowing through it was small and the amounts were offshore-account-sized. Crypto changed the denominator. Broker reporting under the 1099-DA regime is now phasing in, exchange records are being matched against individual returns, and a multi-year inventory of unreported 2017-through-2021 gains sits in the pipeline. The IRS is not discovering a new category of taxpayer. It is discovering that its existing confession booth has one door and no ventilation.
Decompose the 120 days and the reform stops looking like mercy. Pre-clearance review, exam, penalty computation, closing agreement โ four gates, each requiring an agent, each historically consuming weeks of correspondence. The binding constraint was never the statute. It was headcount and perimeter. When I reconstructed the FTX liability gap in November 2022 and arrived at roughly $8 billion in unbacked obligations, I did it in about nine weeks โ not because the work was easy, but because I had fixed the evidentiary boundary in advance. The IRS has the opposite problem: an unbounded perimeter and a bounded number of examiners. You do not fix that with a nicer penalty schedule. You fix it by refusing to let a case stay open.
The 120-day target is a capacity statement disguised as a courtesy. An agency that promises to close in four months is telling you it has a warehouse problem, not a conscience problem.
Which brings us to the arithmetic nobody has published. Take a taxpayer with $400,000 of unreported 2021 gains and a resulting tax deficiency near $95,000. At the 75% civil fraud penalty on the highest year, the disclosure costs him roughly $166,000 before interest. Cut the penalty to 30% and the same disclosure costs about $124,000. Cut it to 10% and it costs $105,000 โ barely above the tax itself. The spread between those three outcomes, roughly $60,000, is the entire incentive the reform is trading on. And the IRS has published none of it. "Significantly lower" is not a rule. It is an unmatured option, and options get exercised by waiting.
Here is the second-order effect that the announcement itself creates. If a rational taxpayer believes the price of confession will fall in December, he does not confess in October. The promise of a cheaper disclosure manufactures a disclosure vacuum in the months before it takes effect โ a self-inflicted dead zone in exactly the window the agency says it wants to shorten. I have watched this pattern before, in a different register: the Curve governance analysis I published in June 2020 showed the same structure. When voting power is decoupled from long-term stake, rational actors defer participation until the incentive curve turns. Governance design and penalty design fail the same way. Code is law until the economy breaks it.
The absence of a digital-asset carve-out deserves more attention than it has received, because it is the only unambiguous decision in the entire announcement. Congress has been building the opposite structure. Section 502 of the Digital Asset Taxation Certainty Act โ reported through the House Ways and Means Committee โ would create a dedicated disclosure channel for taxpayers whose crypto reporting errors were unintentional. That bill encodes a subjective standard: did you mean to underreport? The IRS, by declining a separate track, encodes an objective one: did the return match the ledger?
Those two standards cannot both govern. Intent adjudication is expensive โ it requires interviews, correspondence, judgment calls, and it produces litigation. Throughput is cheap. An agency under inventory pressure will choose throughput every time, and the legislative safe harbour, if it ever reaches the floor, will arrive as an overlay on a machine already running.
Now the contrarian read, which is the one I would stake capital on. A faster, cheaper disclosure channel is not a softening of enforcement. It is an enforcement multiplier. Every generous window in the history of American offshore enforcement followed the same sequence: the 2009 and 2011 programs, the 2014 transition, each one paired with a subsequent widening of the data aperture โ FATCA, then CRS, then automatic exchange. Build the confessional first. Then build the camera network. The confessional is cheap, it generates admissions, and admissions generate the pattern library that makes the camera network worth installing.
Read Koopman's venue choice in that light. He floated this at a tax controversy conference, not in a published revenue procedure. No notice-and-comment period has been described. That is a trial balloon, deliberately permeable, and the details that survive contact with the industry will not be the details announced. The December date is a priority, not a commitment. Administrative systems slip. The penalty number is where the internal argument still lives โ hardliners resisting the discount, pragmatists arguing that a collected dollar beats a litigated one.
What should a practitioner actually do with this? Not file. Wait โ but wait with a calendar, not with hope. The single observable that resolves the entire uncertainty is whether the formal rule text arrives with a percentage attached. If it does, the waiting option collapses, filings surge into the first quarter, and the 120-day clock gets its first real stress test. If December passes and the text still reads "significantly lower," then nothing has been priced, the sophisticated filer keeps deferring, and the IRS continues holding a queue it publicly promised to shorten โ with interest and failure-to-pay penalties quietly accruing against every month of that deferral.
The reform is real. The direction is one-way: the United States is standardising crypto into its ordinary tax machinery rather than granting it a parallel one. What remains unknown is the price of admission. Watch for a number. The number is the policy.