By September 30, federal agencies must publish a full list of their data on their websites. Each entry has to follow a specific, standardized format, so the information is easier for computers and AI tools to read. The requirement comes from a 2019 law called the OPEN Government Data Act, and specific rules for how to do it were laid out in a follow-up memo from the White House’s Office of Management and Budget (OMB), according to a FedScoop report.

To help agencies get ready, the Chief Data Officers (CDO) Council released a 78-page guide in late August. The guide explains a new labeling standard called DCAT-US v3.0, which is a shared set of tags agencies use to describe their data, like a title, a description, and who owns it, written in a consistent way so both people and software can understand it. 

The guide also walks agencies through submitting their inventories to Data.gov, the government’s central website for sharing public data with the public and with other agencies.

“The government cannot exist without data, but you can’t protect it if you don’t know that you have it,” CDO Council Chair Kirsten Dalboe told FedScoop. “AI readiness is nearly impossible if you don’t know what you have and what it can or cannot be used for.”

A Bigger Job Than Before

Agencies have published lists of their public data before. But this law goes further: agencies now also have to document data that can’t be made public, and explain why. That’s a real change for many agencies, Dalboe told FedScoop, since they hadn’t been keeping this kind of record for their private data before.

The guide also connects this work to cybersecurity, including a principle called zero trust. Zero trust means a system doesn’t automatically trust any user or device, and it keeps checking and verifying, even if they are inside the network. Applying that here means agencies have to justify how sensitive each piece of data is and why, not just list it.

Dalboe acknowledged, in the same interview, that not every agency will finish a complete inventory by September 30. The guide recommends doing it in stages instead of trying to finish everything at once.

Why This Matters More in the Age of AI

Taka Ariga, senior director of data policy at the Data Foundation, told FedScoop that one of the guide’s strengths is including actual code examples agencies can copy, not just policy language.

He also explained why getting this right matters more now that AI tools are everywhere. Government terms can be ambiguous. For example, “CDC” could mean the Centers for Disease Control and Prevention, or something else entirely, depending on the agency. 

Without clear labeling, an AI tool might guess wrong and give an inaccurate answer. Done well, Ariga said, this update could make government data far easier for AI tools, like Gemini, ChatGPT, or Claude, to use correctly, without most people ever noticing the difference.

A Deadline Meeting a Shrinking Workforce

The timing adds pressure. A 2025 survey by the Data Foundation found that two-thirds of federal CDOs saw staffing changes over the past year, and nearly 40% lost six or more staff members. Some agencies also lost the software tools they were using to catalog their data, Dalboe told FedScoop, leaving some considering doing the work in spreadsheets instead.

The timeline got tighter too. The new labeling standard was originally expected in July 2025, but wasn’t actually published until May 2026, leaving agencies less time to prepare than planned.

Knowing Our Data Before We Use It

For CDOs outside government, this is a real-world example of what it takes to organize data at scale under a hard deadline, with fewer staff and, in some cases, fewer tools. 

The core lesson applies just as much in the private sector: an organization can’t safely use or protect data it hasn’t identified and labeled first. Federal CDOs are working through that challenge in public right now. Their approach, making steady progress in stages instead of chasing one perfect deadline, is a model any data leader facing a similar cleanup can borrow.