\n\n\n\n Secrecy Is a Terrible Architecture for Public Trust - AgntAI Secrecy Is a Terrible Architecture for Public Trust - AgntAI \n

Secrecy Is a Terrible Architecture for Public Trust

📖 5 min read•863 words•Updated Oct 12, 2026

Picture a Tuesday night in a township meeting room. Folding chairs, a projector that won’t focus, a zoning board of five people who have day jobs. A developer’s attorney stands up to describe a proposed campus: square footage, setbacks, a landscaping buffer. Someone in the third row asks how much water the cooling system will draw in August. The attorney says he isn’t able to discuss that. Someone else asks which company will actually operate the building. Same answer. The room does the math it can do, which is that the people asking for a permit will not say what the permit is for.

That moment, repeated across hundreds of municipalities, is the actual mechanism behind the numbers now circulating. A Gallup poll found 70% to 75% of Americans oppose building large data centers near their homes, and more than half support a national moratorium. Those are not the numbers of a technology that is misunderstood. They are the numbers of a process that reads, from the outside, as deliberately unreadable.

Why an opacity habit became a political problem

I spend my working life on systems whose behavior is hard to inspect. Agent architectures fail in a specific way: when you cannot observe intermediate state, you cannot attribute outcomes to causes, and every stakeholder substitutes a worst-case model for the missing information. Humans do this too. Deny a community the load profile, the water draw, the tenant identity, and the interconnection timeline, and they do not hold the slot empty. They fill it with the most alarming plausible value.

Nondisclosure agreements with local governments were never designed as a public-facing posture. They came out of a competitive logic that made sense in a different era of siting: keep rivals from reading your expansion map, keep land prices from spiking, keep utility negotiations private. The externality is that a township supervisor who has signed one cannot explain to constituents what they approved, or why. The supervisor absorbs the political cost of a decision they are contractually barred from defending. Do that often enough and you do not merely lose a permit. You lose the class of local officials who were previously willing to work with you.

What dropping NDAs would and would not fix

So would ending them persuade opponents? The honest answer, and the reporting on this question does not pretend otherwise, is that nobody knows. It is worth being precise about the two failure modes being conflated.

  • Information failure. People oppose a project because they cannot evaluate it. Disclosure genuinely helps here. Publishing load curves, consumption-based water figures, noise modeling at the property line, and the cost-allocation terms with the utility converts a guessing game into an argument about numbers.
  • Interest failure. People understand the project perfectly and still object, because the benefits land on a balance sheet elsewhere while the ratepayer impact, the turbine hum, and the substation land next door. No amount of transparency resolves that. It only clarifies the disagreement.

Conflating these is the industry’s recurring mistake. Transparency is treated as a persuasion tactic, which means it gets deployed selectively and withdrawn when it stops being convenient. That is self-defeating. Disclosure is only credible when it is structural: the same figures, in the same format, whether they flatter the project or not. A voluntary gesture that evaporates under pressure carries no information, because everyone correctly discounts it.

Regulation arrives when trust does not

The political dimension is the part that should concentrate minds. In Pennsylvania, Governor Josh Shapiro signed an executive order imposing strict standards on data center development. Michael Spencer’s read is that 2026 is the year the backlash against AI data centers accelerated, spreading well beyond the places that had been fighting it first. These two facts belong together. When private negotiation loses legitimacy, the terms get set publicly instead, by people with less interest in your build schedule and more interest in their constituents.

Governors writing standards by executive order is what the absence of workable local process looks like at the next level up. The industry’s choice was never transparency versus speed. It was voluntary disclosure on terms you help design, or mandatory disclosure on terms you do not. The window for the first option narrows every time a township meeting ends with an attorney declining to answer a question about water.

A systems view of the fix

If I were designing this as an engineering problem, I would start from observability rather than messaging. Define a standard disclosure schema for siting: energy, water, noise, tax abatement terms, cost allocation, construction and permanent headcount, decommissioning obligations. Publish it before the hearing, not in response to it. Make the operator’s identity a precondition of the application. Accept that some projects will be rejected on their merits, and treat that as the system working.

Retiring NDAs will not convert an opponent whose objection is substantive. What it does is restore the conditions under which an argument can happen at all. Right now the industry is trying to win a debate while declining to state its position, and losing three-quarters of the room before anyone speaks. That is not a communications problem. It is a design flaw, and it is fixable.

🕒 Published:

🧬
Written by Jake Chen

Deep tech researcher specializing in LLM architectures, agent reasoning, and autonomous systems. MS in Computer Science.

Learn more →
Browse Topics: AI/ML | Applications | Architecture | Machine Learning | Operations
Scroll to Top