When 3% Is the Whole Story
The SEC’s latest enforcement action contains a number that reveals everything: of $12.3 million raised from approximately 150 investors, only $380,000 — about 3% — was ever used to purchase cryptocurrency. The rest was split between personal enrichment and Ponzi-like payments to keep earlier investors quiet.
Texas resident Nathan Fuller, operating through Privvy Investments LLC and Gateway Digital Investments, ran this scheme from at least October 2022 through mid-2024. The pitch was built entirely around “AI”: proprietary bots that could scan for crypto arbitrage opportunities, execute high-frequency trades, and protect capital with automated stop-loss logic. None of it existed.
The Returns That Should Have Ended the Conversation
Promised returns of 40% to 50% within 30 to 45 days — with some investors told to expect more than 100% in under a month — are not a feature of the pitch. They’re the disqualifying condition. Top quantitative funds with institutional infrastructure, decades of data, and deep liquidity consistently target annualized returns in the 20–40% range. A retail operation claiming monthly doubles via automated arbitrage should trigger immediate skepticism regardless of how the pitch is framed.
Fuller’s actual deployment of capital makes the math even starker: $6.2 million went to personal expenses (home purchase, gambling, vehicles, travel); $5.5 million circulated as payments to earlier investors; and the $380,000 that reached crypto markets wasn’t even traded by bots — it was manual, generated no profits, and served only to create the appearance of legitimate activity.
The AI-Generated Paper Trail
One detail in the SEC complaint deserves particular attention for what it signals about the current fraud landscape: when withdrawal requests intensified and the scheme began to unravel, Fuller used AI to generate a letter from a fictitious auditing firm. The letter claimed investor accounts were under review and would eventually be liquidated into a trust.
This isn’t a sophisticated attack vector — it’s a consumer-grade application of freely available AI tools. The implication is straightforward: fabricated documentation is now cheap and accessible enough to feature as a standard component of retail crypto fraud. A professional-looking letter from a “firm” that doesn’t exist costs essentially nothing to produce.
The broader deception package included fabricated account statements showing fictional gains and references to entities that didn’t exist — a pattern that held together long enough to keep investors from withdrawing before Fuller ran out of runway.
A Practical Checklist for Traders
The DOJ’s parallel bankruptcy proceeding — in which Fuller admitted operating a Ponzi scheme and was denied discharge of more than $12.5 million in debt — confirms the pattern: the exit is never clean. Every fabricated document and fake entity becomes evidence.
For traders evaluating any AI-powered trading platform, the Fuller case surfaces four practical filters:
Verifiable performance, not promises. Legitimate platforms publish auditable data — backtests, live performance logs, third-party verification. Fixed or guaranteed return promises at any level should be treated as automatic disqualifiers.
Regulatory standing. Fuller operated unregistered vehicles in violation of federal securities law. Licensed platforms operate within regulatory frameworks that create accountability and recourse.
Traceable operations. Real trading leaves footprints — exchange accounts, transaction records, verifiable custody. A “black box” AI system with no traceable counterparty activity is a red flag regardless of how compelling the interface looks.
Proportional claims. AI-assisted trading can improve execution quality, reduce emotional decision-making, and optimize risk-adjusted returns. It cannot reliably generate 50% monthly gains. Any claim otherwise is selling something other than a trading system.
The “AI” label carries no credibility on its own. The technology is real and genuinely useful in trading contexts — but its presence in a pitch is not evidence that the pitch is legitimate. Fuller’s scheme demonstrates that clearly enough.