News

Dems ask federal agencies for reassurance DOGE isn't feeding data into AI willy-nilly

The Register · Brandon Vigliarolo · last updated

House Democrats have sent letters to 24 federal agencies asking for assurances that Elon Musk’s DOGE team is not feeding sensitive government data into “unapproved and unaccountable” AI systems.

The letters [PDF], penned by Gerald Connolly (D-VA), the most senior Democrat on the House Committee on Oversight and Government Reform, refer to several national newspaper reports claiming the Trump-blessed cost-trimming job-cutting DOGE unit has whipped out commercial AI tools at federal agencies to analyze projects, staffing, and other data without first carrying out appropriate safety and security reviews. The lawmakers say it would be inappropriate for the personal information of Americans to be sifted through by such systems.

Connolly also expressed concern over reported claims DOGE has proposed cuts at many agencies based on the advice of unvetted AI systems following this analysis work.

“Federal agencies are bound by multiple statutory requirements in their use of AI software,” Connolly said, citing Federal Risk and Authorization Management Program (FedRAMP) standards for cloud computing security as one such requirement.

“Failing to ensure that AI vendors … have adequate FedRAMP approval risks violating the Privacy Act of 1974, the Federal Information Security Management Act, and the E-Government Act of 2002,” Connolly continued.

The letter also cited independent analyst Cyberintelligence Brief, which claimed supply chain management platform Inventry.ai is ingesting government data. The analyst claims “multiple US government IP addresses point to its REST API.” The lawmaker’s letters claim the machine-learning outfit has not been approved through FedRAMP.

“These actions demonstrate reckless AI misuse, blatant disregard for data privacy, and a severe failure to maintain the cybersecurity of federal systems,” Connolly wrote.

We’ve asked Inventry.ai for comment.

The letter also states that other federal statutes require all current and planned uses of AI in the federal government to be thoroughly and publicly documented, which the Congressman noted includes all of DOGE’s use of AI.

Citing reports on AI use in the US Department of Education (ED), Connolly claimed that DOGE’s machine-learning toolkit may have been handed datasets containing agency financial data, information on ED employees, and information on student loan borrowers.

“I am deeply concerned that … borrowers’ sensitive information is being handled by secretive members of the DOGE team for unclear purposes and with no safeguards to prevent disclosure or improper, unethical use,” Connolly said.

The Congressman also cited concerns that Musk may be using government data to train his own AI model, Grok, although this seems to us speculative with no evidence to back it up.

The DOGE team didn’t respond to questions asked for this story.

Connolly has given the 24 federal agencies he contacted until March 26 to respond to questions in his letter. Among the requests he made are an explanation of the legal authority Musk and DOGE have used to justify their access to government data, what data has been collected and analyzed by AI, what privacy protections have been deployed to ensure data isn’t exposed, what AI products are being used (including names, algorithm versions, training datasets, questions being asked, records and logs – the works), plus risk assessment records.

“Without clear purpose behind the use of AI, guardrails to ensure appropriate handling of data, and adequate oversight and transparency, the application of AI is dangerous and potentially violates federal law,” Connolly concluded.

The Register has contacted all 24 federal agencies to ask questions about the letter, their responses, and whether DOGE has deployed AIs in their systems. Most didn’t respond, but those that did declined to comment. ®

PS: A small bipartisan group of Congressfolk, including Senator Ron Wyden (D-OR), have urged the UK not to, for the purposes of surveillance, force Apple to disable the end-to-end encryption of certain iCloud data stored at rest. This follows earlier political pressure from the group to change the minds of the British government.

Related Knowledge Center: Artificial intelligence

State of Play

On top of a failing business model, newsrooms must navigate the powerful — and not fully reliable — new capabilities of artificial intelligence and the deployment of large language models.

State of the business model | Copyright and intellectual property


More informationabout Artificial intelligence

Related Knowledge Center: Digital privacy and data security

State of Play

If you’re online, you’re generating a constant stream of information about yourself — sometimes voluntarily, sometimes unwittingly — that is assembled and used to craft your online experience. And generate wealth for data brokers. Meanwhile, designers and managers of the digital infrastructure on which the world depends are engaged in a never-ending effort to thwart attempts, by both criminal enterprises and state actors, to exploit vulnerabilities in those systems.   

Signalgate | State surveillance


More informationabout Digital privacy and data security