Is a team led by Elon Musk going to crash US government computer systems? In recent weeks, there have been numerous reports of Musk and his associates gaining access to servers at departments ranging from the US National Oceanic and Atmospheric Administration (NOAA) and the US Treasury. It isn’t clear if these complex, sensitive and ageing computer systems are being updated, or if private data is being accessed, but there are fears of serious harm being caused.
Musk’s staff are operating under a task force called the Department of Government Efficiency (DOGE), which isn’t actually a government department, but the result of an within days of taking office. Its stated aim is “modernizing Federal technology and software to maximize governmental efficiency and productivity”. Quite what this means in practice is unclear, but experts are concerned about Musk’s seemingly unchecked mandate to access and modify the computer systems that run the US’s most vital institutions.
Reports suggest that software developers , with little or no government experience, have , such as those at the Treasury which process a billion payments annually, totalling $5 trillion. These systems store sensitive information on bank accounts and social security payments. There are also reports that DOGE is to identify targets for budget cuts.
Advertisement
Because these systems have been running for decades, the computer code behind them is often difficult to understand and any changes would normally be approached with caution. “If they are indeed pushing untested code to the live systems, it’s a disaster waiting to happen,” says at the University of Surrey, UK. “You just don’t do it. If it turns out they have the ability to change the systems, and they are doing so with untested code, it’s like fiddling with an aircraft’s engines mid-flight.”
, founder of software company OpusVL and president of The Perl Foundation, says that bold tinkering without the usual safety nets of testing could have serious repercussions, but no outcome is “going to take us back to the Stone Age”.
Such aggressive IT updates happen all the time in Silicon Valley, says Mackintosh, and while the stakes may be higher at the US Treasury than a social media company, any services that are brought offline are unlikely to be terminally broken, although you could easily expose security flaws that leak private data or cause a temporary shutdown, he says.
Not everyone has reacted negatively to DOGE’s approach. Using AI to streamline interactions with old systems could actually bring a great deal of opportunities, making it easier for users to do their work and take informed decisions, says , a retired Microsoft executive and now chairman of trustees for the National Museum of Computing at Bletchley Park, UK.
“We have been gluing new front ends to old applications for decades, especially to offer web and smartphone interfaces for consumers,” says Herbert. But he does have concerns regarding what a new AI front end might do with the data it can now access, particularly personal data and confidential corporate data. “How will we know the AI understands and meets such things as security policies, privacy guidelines and so forth?” says Herbert.
If AI is used to actually change the code of the old systems, with a view to making them inherently more efficient, the stakes are rather higher. If the AI generates code that has bugs or malicious effects, how is this checked before it goes live?” says Herbert.
at Ulster University, UK, says that introducing security flaws to code could even be a boon for adversary governments. “Pushing out untested code can lead to a variety of issues, ranging from minor bugs to catastrophic system failures,” says Curran. “In the case of US national security, foreign hackers might exploit these vulnerabilities, leading to data breaches and system compromises on core US government systems.”
Tinkering with the machinery that underpins the US government is risky, and best done carefully and methodically. So it is little surprise that rushing at the task at pace, while also introducing unpredictable AI models, and doing it without oversight and to opaque ends, is concerning to legal and technological experts.
These concerns are perhaps best summed up by US senator Elizabeth Warren, Scott Bessent, the newly-appointed Secretary of the Treasury: “You appear to have handed over a highly sensitive system responsible for millions of Americans’ private data – and a key function of government – to an unelected billionaire and an unknown number of his unqualified flunkies.”
żěè¶ĚĘÓƵ contacted the US Treasury for clarification but received no response.
