There’s generative artificial intelligence everywhere one looks these days, so much so that it’s impossible to avoid. Every viral video is suspect, teachers and job recruiters lament having to scour submissions for traces of actual intelligence, and more than a few lawyers have been caught citing hallucinated court cases in their legal filings. Not even Congress is exempt from the technological shift. According to The Washington Post, an increasing number of congressional staffers — and members of Congress themselves — are using AI to make their workflow lighter. In doing so, they’re opening a potential Pandora’s box into the already fraught process of crafting America’s laws.
Congress is opening a potential Pandora’s box into the already fraught process of crafting America’s laws.
“Both chambers have cleared members and staff to use chatbots — Copilot, ChatGPT, Gemini and, in the House, Claude — for official work, with more under review,” the Post reported. “The House bought 6,000 Microsoft Copilot licenses last year and encouraged staff to use them.” Roughly half of those licenses are in use with one key guardrail: The AI can’t act on political requests, such as criticizing the opposition party or a specific politician.
(That feels like an easy enough limitation to get around, though. We might expect lawmakers and their staffs to use the same sleight of hand that lets outside groups circumvent campaign finance laws and run “issue ads” that clearly favor one candidate or another.)
There are broad guidelines in place in the House and Senate, the Post reported, including prohibitions on “putting sensitive material such as constituent information into a chatbot, generating deepfakes, making personnel decisions and finalizing legislation.” It’s unclear if any House staffer has been reprimanded for violating those guidelines. Beyond that, both chambers “encourage individual offices to implement their own written AI policies.” The result is a patchwork of hundreds of different standards.
On any given day on Capitol Hill, legislative staffers are busy writing bills that will likely go nowhere, amendments to those few bills that have a chance of clearing both chambers, and the various other bits of writing, like speeches and letters to constituents, that fill their hours.
In that context, the desire for shortcuts is understandable. Legislating is hard work. Legislative text is notoriously fiddly, often citing previously passed laws and amending sections of the U.S. Code. Mistakes could cause massive headaches down the line.
Even in the best of times, there’s plenty of opportunity for human error in the legislative process. A typo in a controversial spending cuts bill caused the House and Senate to pass slightly different versions in 2006, which prompted a potential constitutional crisis. This year, the House scrambled to recall a bill from the Senate that didn’t include an amendment that had passed on the House floor.
Rampant use of AI without proper diligence threatens to make such mistakes more likely. Unfortunately, according to Politico, the House Office of Legislative Counsel, which is charged with catching those errors, has been overwhelmed by a surge of AI-drafted content. The use of AI by congressional staffers and by outside groups in crafting their proposals “is forcing the OLC’s lawyers to spend significantly more time reviewing and rewriting the proposals.” And as Warren Burke, the OLC’s head, testified at a hearing in March, material “produced by outside entities, whether a human or a computer, often takes significantly longer to review and revise as necessary to ensure the text accomplishes the sponsor’s intended policy goals.”
Most lawmakers are already overly reliant on their staffers to translate their policy proposals into legislative text. For those staffers to now be outsourcing that work to chatbots is worrisome. We should all wish more House and Senate members had a firm grasp on the intricacies of legislating and took a more direct role in the process, but Congress appears headed in the opposite direction.
We should all wish more House and Senate members had a firm grasp on the intricacies of legislating and took a more direct role in the process, but Congress appears headed in the opposite direction.
Congressional offices using the technology that they should be regulating may raise concerns for some people, but that isn’t the whole issue. It’s better to have senators and representatives who understand the bare minimum of how today’s internet works. But at the risk of sounding like a Luddite, I’ll express two main fears. The first is that Congress members’ overconfidence in AI’s capabilities will make the already complex tangle of federal law impossible to navigate. Much like software code, inserting too many broken lines into the U.S. Code could cause a cascading failure — and the cost of patching the error could be enormous.
The second, and less tangible, fear is the loss of knowledge and skill this shift portends. Every word, every clause, every punctuation mark in a law is intentional and has meaning. By placing too much trust into chatbots, lawmakers and their staff are making it harder to fully understand what impact the words they make legally binding will have on the people subject to those laws.
The post Congress is getting bogged down by AI slop appeared first on MS NOW.
Author: Source Author
From MS Now.
Leave a Reply