SoylentNews

upstart [soylentnews.org] writes:

████ # This file was generated bot-o-matically! Edit at your own risk. ████

by jan

When it comes to US AI rules, there's too many cooks [theregister.com]:

The US government wants AI in every corner of government, but the unstoppable force of new tech is running into the immovable object of bureaucracy - a growing mass of AI rules.

It's been well established in the first year of Trump's second presidency that AI is a priority [theregister.com] for the administration. Even prior to Trump taking office, government generative AI use cases had surged, growing ninefold [theregister.com] between 2023 and 2024. In recent months, agencies [theregister.com] have cut numerous deals [theregister.com] with most [theregister.com] leading AI companies [theregister.com] under the General Services Administration's Trump-driven OneGov contracting strategy. These agreements give federal agencies access to leading AI models for $1 or less per agency for the first year, suggesting that the Trump team is keen on acting fast.

But given the nature of government, it's not surprising to hear from the auditors at the Government Accountability Office (GAO) that agencies face a large, fragmented set of AI requirements.

According to a report [gao.gov] published on Tuesday, the GAO identified 94 separate "AI-related government-wide requirements" that agencies have to adhere to - and those rules aren't centralized under a single management body.

AI rules and requirements come from ten separate executive-branch oversight and advisory groups, including the Office of Management and Budget, Office of Science and Technology Policy, Department of Commerce, GSA, and National Science Foundation. Those groups help set and police requirements drawn from five AI-related laws, six executive orders, and three guidance documents, the GAO said.

In short, there are a lot of rules and regulations surrounding federal government AI use to account for, making for a tricky - and possibly shifting - deployment path for agencies to navigate.

The GAO declined to take a stance on whether there were too many AI regulations for federal agencies to account for, or whether a central AI regulation for federal use is necessary to streamline operations, but its own prior work suggests that there's a rather severe AI regulatory burden placed on agencies.

It noted [gao.gov] that agencies were struggling with GenAI deployment. Rather than chalk it up to chasing use cases [theregister.com] that don't exist [theregister.com], those agencies pointed to familiar hurdles - a lack of computing resources, concerns over bias and hallucinations, and - you guessed it - too many rules.

Of the 12 agencies the GAO spoke with about AI implementation for that report, 10 said that existing federal AI policy either didn't account for all the obstacles an agency could face when implementing AI or, conversely, that federal AI policy "could present obstacles to the adoption of generative AI."

• Three US agencies get failing grades for not following IT best practices [theregister.com]
• Google, OpenAI, Anthropic get blanket deal to saturate US government with their AI [theregister.com]
• List or get off the pot: Auditors demand gov't improve IT reporting or give it up [theregister.com]
• Perplexity wants to get discounted AI products into the US government too [theregister.com]

"AI is rapidly growing and holds substantial promise for improving the government," GAO's director of IT and cybersecurity Kevin Walsh told The Register in an email. "But AI technologies also pose risks."

Walsh told us that AI can substantially improve operations at federal agencies, but can also be misused to enable cyberattacks, commit fraud, and deanonymize data.

"The rules that govern AI will be critical in our attempts to ensure AI is used for good," Walsh told us while declining to take a position on whether the current fragmented, cumbersome AI regulatory regime was an appropriate one.

That said, the GAO has been making recommendations to improve federal AI oversight as far back as four years ago, when it first published a "framework to help managers ensure accountability and the responsible use of AI in government programs and processes," according to its latest report. In 2023, it issued a second report on how federal agencies were complying with AI rules, but few of the compliance recommendations it made have been acted upon as of yet.

"We made 35 recommendations to 19 agencies … to fully implement federal AI requirements," the GAO said. "As of July 2025, three agencies had implemented four recommendations." ®

Get ourTech Resources [theregister.com]ShareMore about

• AI
• Donald Trump
• Federal government of the United States

More like these×More about

• AI
• Donald Trump
• Federal government of the United States

Narrower topics

• AIOps
• Cybersecurity and Infrastructure Security Agency
• Cybersecurity Information Sharing Act
• DeepSeek
• DOGE
• Federal Aviation Administration
• Gemini
• Google AI
• GPS
• GPT-3
• GPT-4
• Immigration and Nationality Act of 1965
• IRS
• Large Language Model
• Machine Learning
• MCubed
• NASA
• National Highway Traffic Safety Administration
• National Institute of Standards and Technology
• National Labor Relations Board
• NCSAM
• Neural Networks
• NLP
• Retrieval Augmented Generation
• Star Wars
• Telecommunications Act of 1996
• Tensor Processing Unit
• TOPS
• United States Department of Defense
• United States Department of Justice
• US Securities and Exchange Commission

Broader topics

• Government
• Self-driving Car
• United States of America

More about ShareMore about

• AI
• Donald Trump
• Federal government of the United States

More like these×More about

• AI
• Donald Trump
• Federal government of the United States

Narrower topics

• AIOps
• Cybersecurity and Infrastructure Security Agency
• Cybersecurity Information Sharing Act
• DeepSeek
• DOGE
• Federal Aviation Administration
• Gemini
• Google AI
• GPS
• GPT-3
• GPT-4
• Immigration and Nationality Act of 1965
• IRS
• Large Language Model
• Machine Learning
• MCubed
• NASA
• National Highway Traffic Safety Administration
• National Institute of Standards and Technology
• National Labor Relations Board
• NCSAM
• Neural Networks
• NLP
• Retrieval Augmented Generation
• Star Wars
• Telecommunications Act of 1996
• Tensor Processing Unit
• TOPS
• United States Department of Defense
• United States Department of Justice
• US Securities and Exchange Commission

Broader topics

• Government
• Self-driving Car
• United States of America

TIP US OFF

Send us news [theregister.com]

Original Submission