Fifteen million dollars. That’s the price tag for influencing how the U.S. Congress thinks about AI risk. Public First Action, a super PAC, just dropped that sum to back 16 Republican lawmakers who pledge to prioritize AI safety. Seven million has already been spent on ads.
The number stings. Not because it’s big—political money is always big. But because no one outside the PAC knows who wrote the checks. The donors are ghosts. The ad content is black-boxed. The full list of 16 lawmakers? Unpublished. This is a classic capital injection with zero on-chain accountability.
Code is the only law that compiles without mercy. But here, the code is missing entirely.
Context: The AI Safety Lobby Goes Dark
Public First Action is a super PAC—a vehicle designed to collect unlimited money and spend it on political messaging. Their stated goal: electing candidates who take AI safety seriously. That sounds noble until you realize the asymmetry of information. The American public doesn’t know if the funding comes from Anthropic’s founders trying to slow down open-source competitors, or from a military contractor wanting to weaponize AI oversight for procurement advantages.
Super PACs are legally allowed to keep donor identities hidden until periodic FEC filings. That gap creates a span of months where the money moves, ads run, and narrative is shaped—all off-chain. In a bull market for AI fear, this lack of transparency is a vulnerability waiting to be exploited.
I’ve spent the last three years auditing smart contracts and dissecting Layer2 sequencer economics. One thing I’ve learned: when value moves without an auditable trail, the system eventually forks into chaos. Political donations are no different.
Core: How Blockchain Can Audit the AI Safety War
Let’s run a thought experiment. What if Public First Action had used a public blockchain—say, a simple ERC-20 donation tracker on Ethereum—for incoming contributions and ad spend? The contract would record each donor’s address (or an anonymized proxy), the amount, and the timestamp. Ad buys could be executed via smart contracts that release funds only after verifiable delivery proofs from TV stations or digital ad platforms are submitted (e.g., via Chainlink oracles).
Such a system would accomplish three things: - Prevent retroactive influence: If a donor’s identity surfaces months later, the on-chain record shows exactly when they contributed relative to a lawmaker’s vote. Correlation becomes irrefutable. - Detect dark patterns: A sudden spike in donations from a single origin address right before a critical AI safety vote would flash red flags. Any analyst could query the chain and see the anomaly. - Enable composable oversight: DAOs, watchdog groups, even rival PACs could write scripts that automatically alert if funding exceeds certain thresholds or comes from addresses linked to foreign entities. No waiting for FEC filings.
I built a prototype during a hackathon last year: a minimal on-chain donation tracker using a Gnosis Safe module. It took me two weekends. The gas cost was trivial—under $500 for tracking $10M in simulated flows. The real friction isn’t technical, it’s political. Donors don’t want transparency. They want deniability.
But here’s the nuance: full on-chain transparency might violate privacy laws (e.g., state campaign finance disclosure exceptions). A compromise is a commitment scheme where donors deposit via a mixer or zk-rollup, and only the PAC’s treasury manager can reveal the true source upon judicial request. That’s still a massive improvement over the current void.
From my experience dissecting Arbitrum Nitro’s WASM engine, I know that hybrid architectures—part transparent, part private—can work if the trade-offs are explicitly documented. The same applies here: a semi-transparent donation chain paired with legal subpoena capability would balance auditability with legal compliance.
Contrarian: On-Chain Donations Could Backfire
Counter-intuitive take: forcing PACs on-chain might make AI safety lobbying more dangerous, not less.
Consider: if donations are fully transparent, competitors (or hostile foreign actors) can pressure lawmakers by exposing their donor base. A representative who takes money from a “safe AI” PAC could be branded as a shill for a specific company. That weaponizes transparency into partisan ammunition, chilling legitimate policy debate.
Moreover, on-chain records are immutable. A mistake—say a wrong memo field linking a donor to a controversial cause—can’t be deleted. In the current system, FEC filings are eventually corrected. On a blockchain, you’re stuck with the noise.
I’ve witnessed this firsthand while auditing a DeFi protocol that tried to “transparently” track developer compensation. The immutable log accidentally leaked the wallet addresses of employees, leading to targeted phishing attacks. Transparency without proper access control is just a bug dressed as a feature.
Audit reports are hope, not guarantee. The same applies to donation transparency. The fix isn’t full-on-chain indiscriminate exposure; it’s selective disclosure with time locks and consent. For example, donors could authorize public release of their identity only after the election cycle ends, preventing strategic backlash during campaigns.
Takeaway: The Vulnerability Forecast
The $15M from Public First Action will likely succeed in shifting the Overton window on AI safety. But the lack of on-chain audit trails creates a systemic risk: in the next 12–18 months, a major scandal will erupt when it’s revealed that a foreign government or a corporate monopolist secretly funded a “grassroots” AI safety campaign. The public will demand accountability, and the industry will scramble for a retroactive fix.
The question is not whether blockchain can solve political donation opacity. It can—the code is trivial. The question is whether the system’s incumbents, including AI safety advocates, want to compile that law without mercy.