“God View”

In 2014, Uber received well-deserved criticism for having demonstrated a “God View” of its users at parties when it launched in a city. This was a special view of the service available only to admins within the company that would allow them to see and track the ridership of any user. Sadly, this potential for violating user privacy is common at many startups, where large amounts of data are collected on each user and administrators are given wide latitude to view anything they want. Databases are often loosely secured and data from disparate sources are joined together in case it’s useful for marketing. Driven by the rise in web advertising, Silicon Valley has embraced the idea of building up dossiers on individuals to assess their suitability for marketing. Facebook and Google allow for precise ad targeting of granular demographics. Data brokers allow users to purchase vast amounts of commercial data about anybody. And companies like Palantir market systems that consume large amounts of information to make automated decisions and target specific individuals for actions.

Of course, you could theoretically opt out of digital products, install blockers and take other efforts to minimize or eliminate your digital footprint. It’s not easy but it’s possible. But what if you faced the same risk in government? After all, you can’t opt out of filing your taxes. You must provide sensitive PII to get a passport or file a Medicare claim. That situation was exactly the nightmare scenario that the led Congress to create the Privacy Act of 1974. This law laid out some general principles for government to follow to ensure it didn’t create powerful databases to monitor citizens. Government agencies were required to document systems that collect data, the type of data they need and other things they are doing to ensure data is used only in specific ways, retained for minimum periods of time and people have the right to correct their own records. It’s common for outsiders to complain about data silos in government and rules that prevent agencies from sharing data with each other. Those rules are deliberate however to create friction and prevent a “god view” of the American public. Only Congress is supposed to be able to change this - like when they authorized the IRS to share data with the Department of Education via the FUTURE Act, but even that process was explicitly set to take years.

DOGE operates like the Privacy Act doesn’t exist. In early February, there were reports that DOGE had detailed up to 20 staffers within the and they were feeding large amounts of data from those systems into an AI hosted on Azure. There have been similar reports that DOGE was using an on-premises AI system at OPM to process emails received from the Government-Wide Email System (GWES). There hasn’t been any formal verification of these claims, but they seem to have been the precursors for a system that DOGE has been working on since March: an unnamed data lake that is aggregating data from DHS, SSA and HHS including Medicaid data.

In other ominous reports, DOGE staff at agencies demanded they be given elevated access to all systems, whether or not they seemed relevant to the original stated mission of reviewing IT modernization and contracting efforts at agencies. At the time, this just seemed like another manifestation of DOGE’s desire for control, with staff seemingly looking to rack up root access for systems in the same way they collected laptops from different agencies. In some striking examples, Gavin Kliger was granted root access to physical security systems at the DOL and at CFPB. This wpuld presumably allow him to not only track who is coming and going to the building, but potentially also to lock people out or let DOGE allies in. Similarly, DOGE worked at several agencies to get global administrator privileges for SSO systems like Microsoft Entra. This would then allow them to grant and revoke access for any other systems without needing explicit approval.

More recently, DOGE has focused these efforts to use as a tool against immigrants. The first of these efforts has been a project to add social security numbers to the SAVE system at DHS that provides immigration status. The ostensible goal here is to prevent undocumented immigrants from illegally registering to vote in order to commit fraud (there is no evidence this isn’t an exceedingly rare problem), but it also means that for the first time, the US government now has a single API to check if a person is a citizen or not. Legal experts are concerned this system was created without following the Systems of Records Notice (SORN) process mandated by the Privacy Act that requires agencies to publish the data they’re collecting and receive comment from legislators and the public. Data experts are concerned that the system is combining data that wasn’t designed to be used together and subtleties of the data and how it’s updated. As one example, people might register to vote immediately after becoming naturalized citizens, but because that update would take a while to be sent to DHS/USCIS and flow into SAVE, this attempt would be erroneously flagged as fraud. DOGE also has had a few embarrassing examples of misunderstanding SSA data that don’t inspire confidence they will rigorously check false positives before it flags citizens as felons.

DOGE has also been focused on building a master database to track and surveil immigrants within DHS. It’s unclear how much development work DOGE has been doing for this project vs. feeding data into existing data lakes and products from Palantir. There are a lot of unknowns about what data is being pulled into this system, how they are attempting to join wildly disparate data sets, how much this system is being validated or blindly trusted, and how this data collection could be wielded against other populations in America. There is a lot we don’t know yet, but it doesn’t sound good. The following systems are possible data sources for this panopticon: