Crypto AI Fraud Defense: DOJ and SEC Target Fake Trading Algorithms
Fake AI and Phantom Trading Algorithms: The New Center of Federal Crypto Fraud Enforcement
On August 24, 2026, two federal juries in two districts returned guilty verdicts in two separate crypto fraud trials. Both cases turned on the same allegation: the defendant told investors that software was trading or mining cryptocurrency when no such software worked. In Las Vegas, a jury convicted Brent C. Kovar of wire fraud, mail fraud, and money laundering for raising $24 million through Profit Connect, a company that claimed to run artificial intelligence on a supercomputer. In San Francisco, a jury convicted Japheth Dillman of wire fraud and conspiracy for raising nearly $1 million into Block Bits Fund on the strength of a cryptocurrency "autotrader" that, according to the government, never existed. Dillman faced a parallel SEC enforcement action from the day of his arrest in 2022.
Two verdicts on one day sends a message. The government has organized a distinct enforcement theme around what the SEC calls AI washing: marketing an investment with claims about artificial intelligence or automated trading technology that the technology cannot support. The theme now runs from civil settlements against registered advisers to criminal jury trials against founders. Anyone who raised money with a pitch deck describing an algorithm should understand how these two indictments were built, why both agencies have named this conduct a priority, and how these cases are fought at trial.
The technology claim is the charged fraud. In both indictments, the government went beyond alleging misuse of investor funds. It alleged that the description of the technology itself, a "fully functional" autotrader in one case and "proprietary supercomputer AI" in the other, was the material false statement that induced the investments. Every white paper, pitch deck, investor email, and website describing an AI or algorithmic edge is now potential trial evidence.
Historical Conduct, Forward-Looking Priority
Both cases are historical in one sense. The Block Bits conduct ran from June 2017 through August 2018. The Profit Connect conduct ran from late 2017 through July 2021. The indictments and trials came years later. Yet the verdicts landed under enforcement policies, announced by both agencies in 2025, that name this exact conduct as the target going forward.
Start with the SEC. In February 2025, the Commission created the Cyber and Emerging Technologies Unit to replace its Crypto Assets and Cyber Unit. The first priority listed in the announcement: "Fraud committed using emerging technologies, such as artificial intelligence and machine learning." Fraud involving blockchain technology and crypto assets appears on the same list. The unit's stated mission is protecting retail investors. The SEC has stepped back from regulation-by-enforcement theories against crypto platforms, but it built a dedicated unit for exactly the fact pattern in Kovar and Dillman: a false technology claim that separates retail investors from their money. The Commission's 2024 AI washing settlements with investment advisers Delphia and Global Predictions announced the theory. The CETU institutionalized it.
DOJ made the same move. In April 2025, the Deputy Attorney General's "Ending Regulation by Prosecution" memorandum disbanded the National Cryptocurrency Enforcement Team and directed prosecutors away from charging regulatory violations against exchanges and platforms. The memorandum redirected those resources toward one category above all: prosecuting "individuals who victimize digital asset investors." It also instructed prosecutors to avoid charges that require litigating whether a digital asset is a security or commodity, and to reach instead for wire fraud and mail fraud. That instruction describes the Kovar and Dillman prosecutions precisely. Neither indictment contains a securities count. Both rest on 18 U.S.C. § 1343 and its companion statutes, which apply regardless of how the underlying asset is classified.
Retail losses drive case selection under both policies. The SEC's 2021 emergency action against Profit Connect alleged that the company encouraged investors to draw on retirement funds and home equity and targeted people saving for their children's education. The Block Bits case involved just over 20 individual investors and roughly $960,000. That is a small number by crypto standards. The government tried the case anyway, through a superseding indictment and a ten-day jury trial. Prosecutors are plainly willing to invest trial resources in modest-loss cases when the victims are retail investors and the technology claim is provably false. Expect the next wave of indictments to follow the same template.
The Block Bits Indictment: Charging a Phantom Autotrader
The superseding indictment in United States v. Dillman, No. 3:23-cr-00140 (N.D. Cal.), charged one count of wire fraud conspiracy under 18 U.S.C. § 1349, four counts of wire fraud, aiding and abetting, and a forfeiture allegation. The core allegation was simple. Dillman and a co-founder told investors that Block Bits Fund would profit through a cryptocurrency "autotrader," an algorithm executing automated arbitrage trades across exchanges. They repeatedly described the autotrader as fully functional. The grand jury alleged that no functioning autotrader ever existed.
The indictment built falsity from specific documents. A pitch deck claimed the fund had "seen incredible increase in the performance of the auto-trader over letting the currency sit or be managed by hand." A June 2017 email to a prospective investor claimed the autotrader had completed five trades over 17 days and achieved a 72.44% growth rate. The government also alleged a second technology misrepresentation: that investor funds sat in safe "cold storage" deals generating returns, when the funds were actually deployed into undisclosed, high-risk crypto ventures. "Cold storage" is a term of art in the crypto community for secure, offline custody. The indictment treated the misuse of that term as its own deception.
"We have seen incredible increase in the performance of the auto-trader over letting the currency sit or be managed by hand."
— Block Bits pitch deck, as quoted in the superseding indictment, United States v. Dillman, No. 3:23-cr-00140 (N.D. Cal.)The most unusual feature of the indictment is its selection of wires. Three of the four substantive wire fraud counts are payments to victims: transfers of $29,970, $50,000, and $24,970 from the fund's bank account to individual investors. The government's theory was that these were lulling payments. They made investors believe the autotrader was generating profits, when the money actually came from other victims and from manual trading. Under this theory, each repayment to a victim became its own felony count. The fourth count was a single August 2018 "Fund Update" email in which Dillman called the autotrader "effective" and told investors a second fund was coming to "take advantage of our automated predictive trading system."
The jury convicted Dillman of wire fraud and conspiracy on August 24, 2026, after a ten-day trial before Judge Richard Seeborg. Sentencing is set for December 8, 2026. Each count carries a statutory maximum of 20 years. His co-founder, David Mata, was charged separately by information, the procedural posture that typically signals a plea and cooperation.
The Profit Connect Indictment: A Supercomputer AI That Guaranteed Returns
The indictment in United States v. Kovar, No. 2:25-cr-00028 (D. Nev.), charged twelve counts of wire fraud, three counts of mail fraud under 18 U.S.C. § 1341, and three counts of money laundering under 18 U.S.C. § 1957. The grand jury alleged that Kovar's company, Profit Connect, purported to use artificial intelligence software on a supercomputer to mine cryptocurrency and verify transactions, and that Kovar promised fixed returns of 15% to 30% APR with a 100% money-back guarantee. He allegedly obtained approximately $24 million from at least 400 investors.
The AI claim sat at the center of the pleading. Profit Connect's website allegedly told investors the company used its "proprietary supercomputer AI to guide the four (4) Wealth Services income streams," the first of which was a "block-chain AI prediction algorithm." The indictment alleged the company was never profitable, had no reserves, and could not pay the promised returns. Investor money instead operated the company, bought gifts for employees, bought Kovar a house, and repaid earlier investors in Ponzi fashion.
Several features distinguish this indictment. First, vocabulary. The grand jury alleged that Kovar deliberately avoided the words "investment" and "investor," calling the investments "seat time" on the supercomputer and the investors "customers" and "depositors." The government pleaded his word choices as evidence of concealment. Second, fabricated corroboration. The indictment alleged Kovar forged a Bank of America letter stating that Profit Connect's accounts were FDIC insured and that the company held "Platinum Honors Client" status, and caused an attorney to draft letters claiming the company held over $700 million, in one version, and over $445 million, in another, in cryptocurrency reserves. Third, real hardware. Kovar leased a warehouse near the Las Vegas Speedway, built computers, and installed a cooling system. The machines existed. The charged misrepresentation concerned what they produced: Kovar allegedly told investors the computers were functioning and generating substantial mining profits when they were idle. Fourth, real-time deception. The indictment quoted a text to sales agents claiming "the largest cryptocurrency verification in our history ... 870,000,000 ... Australia Treasury sending to Beijing," complete with a fabricated verification fee.
The jury convicted Kovar on August 24, 2026, after a nine-day trial, on eleven wire fraud counts, two mail fraud counts, and two money laundering counts. Sentencing is set for November 30, 2026, with a combined statutory maximum of 280 years. The criminal verdict landed five years after the SEC's emergency action froze the company's assets in July 2021.
What Comes Next
The government now has a tested playbook: identify the technology claim, prove the technology performed nothing like the description, trace investor money away from the represented use, and charge the communications and transfers as wires and mailings. The playbook works whether the venture raised $960,000 or $24 million, and whether the technology was pure fiction or real hardware that underdelivered. The SEC opens with an emergency action and asset freeze. DOJ follows with an indictment. Both agencies cite retail victims in every filing, and both have published policy statements committing resources to this exact category of case.
The exposure reaches beyond outright schemes. Founders raise money on roadmaps. Pitch decks describe capabilities in the present tense that engineering treats as aspirational. Marketing teams write "our AI" when the product runs on rules-based scripts or third-party models. In the government's frame, each of those statements is a potential count. The distance between optimism and fraud is measured by three elements the government must prove beyond a reasonable doubt: falsity, materiality, and intent to defraud. Those elements are contested at trial, witness by witness and document by document.
How These Cases Are Fought at Trial
Kovar and Dillman were both convicted, and those verdicts confirm how dangerous this fact pattern is in front of a jury. They also confirm something else: these cases get tried. The government took a $960,000 loss case through a ten-day jury trial. Defendants facing AI misrepresentation charges need counsel who has actually tried federal fraud cases to verdict, because the outcome turns on trial craft, on cross-examination of cooperators and victims, and on making a jury believe in the defendant's good faith. The partners at Armstrong & Bradylyons spent years trying these cases for DOJ before building the firm's crypto investment fraud defense practice. Scott Armstrong served as lead trial counsel in the first cryptocurrency market manipulation case charged under Title 15, a multi-week federal jury trial built on blockchain tracing and exchange data involving more than $300 million in trades. He and Drew Bradylyons know how the government constructs these trials because they constructed them.
The firm defends individuals in federal crypto fraud cases nationwide. Crypto prosecutions concentrate in a handful of districts: the Southern and Eastern Districts of New York, the Northern and Central Districts of California, the District of Nevada, the Southern District of Florida, the Eastern District of Virginia, and the District of Columbia. The Kovar and Dillman trials came out of Nevada and the Northern District of California, two of the busiest venues for digital asset cases. Armstrong & Bradylyons appears in every federal district where DOJ brings these charges, and its partners tried federal fraud cases across the country as prosecutors before crossing to the defense.
Cross-Examining the Cooperator
Nearly every crypto fraud trial features a cooperating co-founder or insider. The Dillman case followed the standard architecture: the co-founder, David Mata, was charged separately by information, and Dillman went to trial alone. A cooperator gives the government its narrator, the witness who translates emails and bank records into a story of shared intent. The cooperator is also the most vulnerable witness in the courtroom. He testifies under a plea agreement, with his sentence hanging on the government's satisfaction. He has told his story through dozens of proffer sessions, and every session generates notes that can be mined for shifts, additions, and coached refinements. Former prosecutors know how cooperators are prepared because they prepared them. That experience shows the defense where the preparation seams are: the first proffer that omitted the key admission, the memory that improved as sentencing approached, the conduct the cooperator minimized to protect his own deal. A jury that sees the cooperator's incentives clearly will discount his account of what the defendant knew.
Handling Victim Testimony
Victim witnesses carry the government's materiality proof, and they carry emotional weight. Attacking them head-on backfires. The effective cross is respectful and surgical, and it aims at a different target: what actually drove the investment decision. Crypto investors in 2017 and 2021 were chasing an asset class in the middle of historic run-ups. Many signed subscription documents disclosing total-loss risk, held other speculative positions, and invested within days of first contact. Each of those facts, drawn out gently from the government's own witnesses, undercuts the claim that the specific algorithm representation moved the money. Materiality must be proven for every count, and it is often the least developed part of the government's case. Trial counsel who has presented victim testimony for the government knows which questions those witnesses have never been asked.
Turning the Documents Against the Case
These prosecutions are built on documents: years of emails, chat logs, pitch decks, bank statements, exchange records, and blockchain data. Volume is the government's weakness as much as its weapon. To present a document case at trial, prosecutors must compress hundreds of thousands of pages into summary charts and a handful of highlighted exhibits, sponsored by case agents and forensic accountants. Every summary chart is a choice about what to leave out, and every omission is cross-examination material. The email with the damaging sentence sits in a thread that also contains the engineering update, the disclosure of a setback, or the defendant's own money going into the venture. Defense counsel who has built government summary exhibits under Rule 1006 knows exactly where the compression distorts, and forces the sponsoring witness to concede, document by document, what the chart conceals. Jurors notice when the full record reads differently from the government's excerpt.
Making the Good Faith Case
Good faith is a complete defense to wire fraud, and it cannot be argued as an afterthought. It has to be the defense theory from the first minute of opening: this defendant believed the technology, spent the money trying to build it, and said what he believed to be true. The proof is affirmative. Engineering payroll and development contracts. Personal funds invested alongside clients. Internal messages candidly discussing setbacks, which show a founder wrestling with reality rather than scripting a lie. The government answers with lulling payments, forged letters, and personal spending, which is why the Kovar and Dillman records were so difficult. Presenting good faith persuasively, deciding whether the defendant testifies, sequencing the defense exhibits so the jury reads the record in context, and arguing absence of intent in closing without conceding an inch on the other elements: that is trial judgment, and it comes only from standing up in federal court and doing it. The firm applies that judgment, with retained blockchain forensics and software experts, in every crypto fraud and money laundering case it defends.
Frequently Asked Questions
What is AI washing in federal fraud and securities enforcement?
AI washing is marketing an investment, product, or company with claims about artificial intelligence capabilities that the underlying technology cannot support. The SEC brought its first AI washing enforcement actions in March 2024 against investment advisers Delphia and Global Predictions, which paid $400,000 in combined civil penalties for overstating their use of AI. In February 2025, the SEC made "fraud committed using emerging technologies, such as artificial intelligence and machine learning" the first listed priority of its new Cyber and Emerging Technologies Unit.
DOJ charges the same conduct as wire fraud and mail fraud when false AI or algorithm claims induce investments. In August 2026, federal juries convicted Brent Kovar, who claimed a "proprietary supercomputer AI" mined cryptocurrency through his company Profit Connect, and Japheth Dillman, who claimed his Block Bits Fund ran a fully functional crypto "autotrader." Armstrong & Bradylyons defends founders, executives, and promoters facing AI washing allegations through its crypto investment fraud defense practice.
What are the penalties for wire fraud in a cryptocurrency case under 18 U.S.C. § 1343?
Wire fraud under 18 U.S.C. § 1343 carries up to 20 years in prison and a $250,000 fine per count, plus restitution and forfeiture of proceeds. Conspiracy to commit wire fraud under § 1349 carries the same maximum. Each charged email, wire transfer, or mailing is a separate count, so exposure compounds quickly. Brent Kovar's fifteen counts of conviction in the Profit Connect case carry a combined statutory maximum of 280 years.
The advisory Sentencing Guidelines range in these cases is driven principally by the loss amount and the number of victims, so a $24 million scheme with 400 investors produces severe exposure even for a first offender. Money laundering counts under § 1957, which attach to transactions over $10,000 in fraud proceeds, add further counts and forfeiture reach. The attorneys at Armstrong & Bradylyons charged and tried these statutes for the government and now litigate loss amount, victim counts, and forfeiture for the defense.
How does the government prove that statements about a trading algorithm or AI system were false?
Prosecutors compare the charged statements against the actual state of the technology at the time each statement was made. In the Block Bits case, the government pointed to a pitch deck describing autotrader performance and an email claiming a 72.44% growth rate over five trades, then presented evidence that no functioning autotrader ever existed. In the Profit Connect case, the government contrasted website claims of a profitable "supercomputer AI" with financial records showing the company had no revenue source other than new investors.
The proof combines bank records, blockchain tracing, developer and employee testimony, internal messages, and forensic analysis of the software itself. Each category can be contested. Tracing rests on clustering heuristics and attribution inferences. Employee witnesses often saw only part of the development effort. Software that failed in production may still have functioned in testing when the statements were made. The defense case on falsity is built from version histories, commit logs, test records, and the dated documents the government's timeline overlooks.
Are optimistic projections about developing technology considered federal fraud?
No. Wire fraud requires a knowingly false statement of existing fact made with intent to defraud. Honest optimism, projections, and aspirational statements about what technology will do fall outside the statute, so long as the speaker believed them. Courts also recognize that vague promotional superlatives can constitute non-actionable puffery. A forecast about future capability sits on one side of the line the government drew in these indictments. A specific claim of completed performance, such as Block Bits' asserted 72.44% growth rate from a working autotrader, sits on the other.
The dividing question is what the speaker knew and believed on the date of the statement. Contemporaneous evidence of engineering work, spending on development, and internal belief that the system functioned bears on both falsity and intent. Armstrong & Bradylyons builds that record in every crypto investment fraud case it defends.
Why did prosecutors charge payments to investors as wire fraud counts in the Block Bits case?
Three of the four substantive wire fraud counts in the Dillman superseding indictment were electronic payments from the fund's bank account to individual victims. The government's theory was that these were lulling payments: transfers designed to make investors believe the autotrader was generating profits, when the money actually came from other investors and from manual trading. Under wire fraud doctrine, a transmission that lulls victims into a false sense of security and postpones discovery of the scheme counts as a wire "for the purpose of executing" the scheme, even though the money moved toward the victim.
The charging choice carries a practical lesson. Repaying investors does not close a fraud investigation, and under a lulling theory each repayment can become its own felony count. The defense answer is to contest the characterization: payments reflecting a genuine, transparent attempt to return capital, made without false profit reporting, fall outside the lulling theory. The dispute turns on what the accompanying communications said and what the defendant intended, which is why the contemporaneous record decides these cases.
How do parallel SEC and DOJ proceedings work in crypto AI fraud cases?
Both August 2026 convictions followed parallel SEC civil actions. The SEC filed an emergency action against Profit Connect in July 2021, obtaining a temporary restraining order and asset freeze nearly four years before the criminal indictment. The SEC charged the Block Bits entities and founders the same week DOJ announced the criminal complaint, and the SEC's San Francisco office assisted the criminal investigation.
Parallel proceedings create acute risk. Testimony, sworn statements, and documents produced in the SEC matter can be used by prosecutors in the criminal case, while asserting the Fifth Amendment in the civil case can support an adverse inference. Asset freezes and receiverships also constrain the resources available for defense. Coordinating both tracks from the outset, including the sequencing of any testimony, is central to how Armstrong & Bradylyons handles matters through its crypto fraud and money laundering practice.
How do defense attorneys challenge cooperator and victim testimony in a federal crypto fraud trial?
Cooperators testify under plea agreements that tie their sentences to the government's satisfaction, and their accounts evolve across proffer sessions whose notes are discoverable. Effective cross-examination reconstructs that evolution: the first proffer that omitted the key admission, the memory that improved as sentencing approached, the conduct the cooperator minimized to protect his own deal. In the Block Bits case, co-founder David Mata was charged separately by information, the posture that typically signals cooperation, leaving Dillman to face the cooperator architecture at trial.
Victim witnesses require a different approach. Attacking them alienates jurors. The productive cross draws out, respectfully, what actually moved the investment: the market run-up, signed risk disclosures acknowledging possible total loss, other speculative holdings, and decisions made within days of first contact. Those admissions undercut materiality, an element the government must prove for every count. The partners at Armstrong & Bradylyons prepared cooperators and presented victim testimony for years as DOJ prosecutors, and they cross-examine both categories of witness with that inside knowledge.
What experience does Armstrong & Bradylyons have trying federal cryptocurrency fraud cases?
The firm's partners spent years as federal prosecutors charging and trying complex financial crime cases before moving to the defense. Scott Armstrong served as lead trial counsel in the first cryptocurrency market manipulation case charged under Title 15, a multi-week federal jury trial built on blockchain tracing and exchange data involving more than $300 million in trades. That trial covered the exact proof the government used against Kovar and Dillman: tracing analyses, exchange records, cooperator and victim testimony, and circumstantial intent evidence assembled from documents.
Armstrong & Bradylyons defends founders, executives, promoters, and traders in federal investigations and prosecutions nationwide through its crypto investment fraud defense and crypto fraud and money laundering practices. The firm appears in every federal district, including the venues where crypto prosecutions concentrate: the Southern and Eastern Districts of New York, the Northern and Central Districts of California, the District of Nevada, the Southern District of Florida, the Eastern District of Virginia, and the District of Columbia. It retains blockchain forensics and software experts to test the government's tracing methodology, wallet attribution, and technology narrative, and it tries these cases to verdict when the evidence supports a defense on falsity, materiality, or intent to defraud.
Facing a Crypto Fraud Investigation Built on AI or Algorithm Claims?
Armstrong & Bradylyons PLLC defends founders, executives, and promoters in federal crypto fraud investigations and trials nationwide.
