Civic
Publishing government data was supposed to change everything. A growing body of evidence suggests disclosure alone was never the mechanism reformers thought it was.
NewsOnScale Staff
July 12, 2026
There is a foundational myth embedded in the civic technology movement, one so widely shared it became invisible: that disclosure is itself a form of power. Open data portals, public spending dashboards, machine-readable budget filings — the infrastructure of government transparency was built on the assumption that making information available was functionally equivalent to making it actionable. A wave of serious institutional research is now challenging that assumption directly, and the implications reach well beyond civic tech into every conversation currently happening about AI, democracy, and the future of public accountability.
The core critique is not that transparency is useless. It is that transparency, as typically implemented, is radically incomplete as an anticorruption strategy. Publishing a dataset does not create the interpretive capacity, the political will, or the institutional pressure required to act on what that dataset reveals. Without those downstream conditions, disclosure functions more like a bureaucratic compliance ritual than a genuine accountability mechanism.
## The Capacity Gap No One Wanted to Talk About
Civic technologists spent enormous energy on the supply side of the transparency equation — building tools, filing FOIA requests, standardizing data formats, pressuring agencies to publish in machine-readable form. Far less energy went into understanding demand: who actually uses this information, what they can do with it, and what structural barriers prevent most people from doing anything at all.
The gap is stark. A government agency that publishes its contracting data in a well-formatted CSV file has technically satisfied its disclosure obligation. But if the journalists who might analyze it have been laid off, if the civil society organizations equipped to interpret it are underfunded, and if the communities most affected by those contracts lack broadband access or data literacy resources, the published file changes nothing. The information exists. Power does not shift.
This is not a peripheral concern for a niche academic audience. It is directly relevant to the current moment, in which AI is being positioned — including by well-funded philanthropic initiatives — as the tool that will finally close the civic transparency gap. The argument goes that if human capacity to process government data has been the limiting factor, machine intelligence can substitute for it at scale.
## AI Doesn't Fix a Political Problem
That argument deserves serious scrutiny before another generation of civic infrastructure gets built on top of it. AI systems can absolutely accelerate the extraction of patterns from large document sets, flag anomalies in financial disclosures, or synthesize public meeting records in ways no human researcher could match at volume. These are real capabilities with real applications in the accountability space.
But the lesson from two decades of transparency work is that the bottleneck was rarely information processing. It was power. Corrupt systems do not generally persist because their participants are successfully hiding the evidence. They persist because the institutions responsible for acting on disclosed information — legislatures, prosecutors, electoral bodies, a functioning press — lack the will, the independence, or the protection to do so.
No language model changes that calculus on its own. An AI that can read ten thousand budget documents cannot compel a city council to respond to what those documents reveal. It cannot protect a whistleblower. It cannot rebuild the local news infrastructure that once created reputational consequences for documented misconduct.
## What Evidence-Based Reform Actually Requires
The most useful reframe emerging from this body of work is a shift from transparency as an end to transparency as a precondition — one that requires explicit investment in the social and institutional infrastructure that converts disclosed information into changed behavior.
That means funding the intermediaries: investigative journalists, civic data analysts, community advocates, and legal aid organizations that sit between raw disclosure and public accountability. It means designing AI tools not just to surface information but to serve the specific workflows of the people most positioned to act on it. And it means being honest about what technology cannot substitute for, which is the painstaking, often dangerous, deeply human work of holding power to account.
The transparency era produced important infrastructure. The accountability era requires something harder to build and harder to fund: the conditions under which disclosed information actually matters.