{"slug": "what-congress-actually-needs-to-do-about-ai-staff-up", "title": "What Congress Actually Needs to Do About AI: Staff Up", "summary": "The Revolving Door Project argues that Congress lacks independent technical expertise to regulate frontier AI, citing the 1995 closure of the Office of Technology Assessment under then-House Speaker Newt Gingrich and the current ratio of one tech lobbyist for every 1.5 members of Congress employed by the six largest technology companies. The piece follows Anthropic employee Jacob Coxon's September 8 public resignation calling for a nationwide AI development slowdown, which Anthropic CEO Dario Amodei endorsed with a three-part pacing proposal involving an antitrust exemption and full-time auditors modeled on U.S. banking regulation, and which OpenAI CEO Sam Altman and Elon Musk also backed. House Democrats are pushing for immediate legislation while a bipartisan Senate trio negotiates a safety bill that would preempt existing state AI laws, a longtime industry goal.", "body_md": "*The Revolving Door Project, a* Prospect *partner, scrutinizes the executive branch and presidential power. Follow them at [therevolvingdoorproject.org](https://therevolvingdoorproject.org/).*\n\nIt wasn’t so long ago that when a scary new technology burst onto the scene and raised uncomfortable new social questions, lawmakers knew just what to do: call their in-house tech think tank to get some perspective.\n\nIn 1974, Congress established the Office of Technology Assessment, a nonpartisan advisory body that provided independent analysis of new scientific and technological issues, from acid rain to polygraph test accuracy. The OTA made sure that if lawmakers were curious about a nascent technology, they had a good source of information besides the industry itself. It was educational, nonideological, and a resounding success.\n\nThis, of course, infuriated the conservative movement. As part of the 1994 “Contract with America,” then-House Speaker Newt Gingrich shuttered the OTA the following year, striking a crucial blow for ignorance, bias, and deference to private industry. Today, six of the largest technology companies collectively [employ](https://issueone.org/press/big-tech-spends-millions-to-buy-influence-in-washington-in-first-half-of-2026/) one lobbyist for every 1.5 members of Congress. Layer on the gutting of local newsrooms due to a Google- and Meta-dominated ad market, plus a handful of [legacy outlets](https://podcasts.apple.com/us/podcast/the-assassination-of-the-washington-post-by/id1651876897?i=1000774754017) now [reshaped](https://www.theguardian.com/media/2024/dec/18/la-times-patrick-soon-shiong-trump) by [tech billionaires](https://www.forbes.com/sites/andymeek/2026/07/07/all-the-companies-and-brands-paramounts-david-ellison-is-about-to-control/) with their own [agendas](https://www.newyorker.com/magazine/2026/01/26/inside-bari-weisss-hostile-takeover-of-cbs-news), and Congress has very few trustworthy sources of information.\n\n[Read more from the Revolving Door Project](https://prospect.org/tag/revolving-door-project/)\n\nThe result is a Congress that is frequently befuddled about technical matters, a problem that has seldom been more obvious than when it comes to so-called frontier artificial intelligence products. On September 8, an Anthropic employee named Jacob Coxon [publicly resigned](https://www.wsj.com/tech/ai/anthropic-researcher-quits-over-out-of-control-ai-fears-707b7628) and called for a nationwide “slowdown” in the development of these AI models, claiming that these products could wipe out humankind by the end of the decade if we aren’t careful.\n\nSounds bad! Maybe Congress should step in and put the kibosh on this whole endeavor? Or is the whole thing just a publicity stunt to drive valuations even further? As it stands, Congress isn’t really equipped to answer these questions. Apparently, the only one who knows how to organize and regulate AI … is the AI industry itself.\n\nWithin days of Coxon’s prognosis, his ex-boss Dario Amodei endorsed Coxon’s worries and [proposed a three-part process](https://darioamodei.com/post/we-must-pace-the-frontier) for slowing development, involving a special exemption from antitrust law and full-time auditors modeled after U.S. banking regulation. OpenAI CEO Sam Altman co-signed the need for pacing and pledged to match Anthropic’s auditor commitment. Elon Musk offered three words on X: “Dario is right.”\n\nThe CEOs found a ready audience on Capitol Hill. As [the](https://prospect.org/2026/09/15/artificial-intelligence-open-ai-anthropic-congressional-democrats/) *Prospect*[reported](https://prospect.org/2026/09/15/artificial-intelligence-open-ai-anthropic-congressional-democrats/), House Democrats are clamoring for immediate legislation while a bipartisan Senate trio is negotiating a safety bill that would preempt the handful of state laws currently on the books, a longtime industry goal.\n\nThe whole situation is strange on its face. Booming industries do not typically beg for government oversight. But once the government hooks market incumbents into a bespoke regulatory system, it becomes a lot harder for regulators to just let them fail in a downturn like any other company. By acting without careful *independent* analysis, Congress risks ensuring the AI inevitability it is so worried about.\n\nWhat’s causing this latest round of panic anyway? What shocked Coxon was the so-called “Hugging Face incident,” in which OpenAI researchers asked their latest models to solve some hacking problems, but left an unguarded entry point to the open internet in the testing area. The models used that entry point to try to break into a third-party server where the problem answers were stored. In other words, OpenAI committed the cybersecurity equivalent of leaving a feral dog unsecured and unsupervised in the backyard, with the neighbor’s trash can full of table scraps just over the fence. To the [limited extent](https://www.yalejreg.com/nc/who-investigates-ais-cyber-attacks/) the public knows exactly what happened, [some experts](https://bsky.app/profile/jmiers230.bsky.social/post/3mvfocpbct223) see this media blitz as the industry [spinning](https://doctorow.medium.com/https-pluralistic-net-2026-09-12-god-in-the-box-llms-are-fake-5d40c669291e) what should be a public embarrassment about OpenAI’s bush-league information security into a government-mandated incumbency advantage.\n\nThe cybersecurity implications of AI are vast, and many important firms are known to underinvest in cybersecurity. But assessing longer-term cybersecurity implications of agentic AI requires significant research and assessment. It seems likely that the answer looks more like setting standards for other businesses while ensuring courts and attorneys general hold AI businesses liable for foreseeable uses of their technology—including enforcing existing laws about hacking and cyber attacks—not granting exemptions to AI developers. But such a conclusion is unlikely if Congress relies for expertise on the firms themselves.\n\nAmodei’s comparison to banking regulation also raises eyebrows. One of the big lessons from that field is how ineffective self-regulation of this sort [can be](https://www.warren.senate.gov/newsroom/press-releases/warren-raises-concerns-about-steep-decline-in-enforcement-by-financial-industrys-self-regulatory-agency/). Moreover, we built a bespoke regulatory system for the banking industry because it is critical social infrastructure—a modern economy cannot function without money. But economies functioned just fine without AI agents less than a decade ago, even if more students were forced to write their own essays. Producing fewer AI agents isn’t really the same as cutting off consumer credit.\n\nAs for the antitrust exemption, former Federal Trade Commissioner Alvaro Bedoya [pointed out on LinkedIn](https://lnkd.in/p/gtPT_WtF) that AI companies are already allowed to share legitimate cybersecurity information with each other under [long-standing guidelines](https://www.justice.gov/archives/opa/pr/justice-department-federal-trade-commission-issue-antitrust-policy-statement-sharing). Bedoya’s colleague Lina Khan also [indicated](https://x.com/linamkhan/status/2099204390548639960) that AI companies may already be liable for consumer protection harms under the good old, general-purpose Federal Trade Commission Act. Locking AI into its own regulatory regime, designed by the industry itself, could make it harder for other regulators to assert jurisdiction, either for fear of stepping on toes or due to hard-coded legal barriers. And tech upstarts from Microsoft to Meta always claim that the antitrust laws do not apply to their special innovations. That never holds up legally or practically.\n\nLet’s also not forget all of that data center debt set to come due in the next few years. Barring some productivity miracle, general-purpose LLMs may be [the money-losingest venture in history](https://www.technologyreview.com/2026/09/15/1144028/ai-infrastructure-boom-investment-bubble-risk/). Stuck with overvalued products, a monstrously expensive capital investment race, and no path to the dizzying profits promised, frontier developers seem to be hoping Uncle Sam will somehow bail them out from their own mistakes.\n\nIn short, there is plenty of reason to be skeptical of opportunistic demands associated with the AI panic, even as there may be real causes for alarm. AI clearly does raise new questions across a huge breadth of policy topics—cybersecurity, consumer protection, financial stability, intellectual property, and so on. Probably no one person can speak with authority and independence on every single one of these topics.\n\nCertainly not the CEOs. A [Ronan Farrow and Andrew Marantz exposé](https://www.newyorker.com/magazine/2026/04/13/sam-altman-may-control-our-future-can-he-be-trusted) in April cited former OpenAI employees saying Sam Altman can barely code and misunderstands core machine learning concepts. He’s not a science whiz; he’s a businessman with a noted history of dishonesty, who’s currently spinning a lot of investment-intensive plates.\n\nThis is why Congress should develop in-house science whizzes before throwing the federal government’s weight at whatever the businessmen say they’d like. Whether it’s drawing up new regulation, or enforcing existing laws, Congress will struggle to control AI or any tech company without any form of in-house advice.\n\nIt’s a problem that goes beyond the lamented Office of Technology Assessment. In Congress and everywhere else, brain drain is always ultimately a money story. According to a [2020 New America study](https://www.newamerica.org/insights/congressional-brain-drain/executive-summary/), salaries for Hill staffers have been falling across the board since before the Great Recession, with most congressional aides staying only about three years. Many aides enter [with the express plan](https://www.newamerica.org/insights/congressional-brain-drain/executive-summary/) of revolving out to lobby on K Street.\n\nThe result is that most of the people trusted to provide expert policy advice to our nation’s legislators are twenty-somethings with barely a year or two of experience, juggling far more issue areas than anyone can reasonably gain proficiency in, while earning barely enough to keep the lights on in a studio apartment. No wonder they turn to the armies of corporate lobbyists; they’re the only ones offering ready answers and assistance with this impossible job.\n\nIndeed, AI technology itself is already affecting Congress’s normal workflow, and not for the better. Politico [reported in August](https://www.politico.com/news/2026/08/17/ai-slop-lawmakers-congress-01008376) that the House Office of the Legislative Counsel—the lawyers who take members’ plain-language policy ideas and translate them into bill text—is facing “a flood of legislative proposals generated using artificial intelligence tools, many of them riddled with errors and shoddy language.”\n\nLegislative drafting is a niche subspecialty in law that requires enormous attention to detail, and whose any error can have huge ramifications. All of the time that drafters spend cleaning up slop proposals from LLMs that can’t distinguish between a tax credit and a tax deduction is time they aren’t drafting the bills members actually ask of them. This could feed into the existing perception that the legislative counsel is a time-wasting exercise that ought to be automated away. Before long, you could have AI-drafted laws referring to statutes that don’t exist.\n\nIn short, if Congress wants to understand AI, it needs more human beings. More member staff, more legislative drafters, and something like a reborn Office of Technology Assessment should all be on the table. Congress needs to finally redress the self-inflicted wounds to its own intellectual capacity, one of the many foolish decisions that have left the legislative branch practically dormant for the last several decades. If the AI boom is what finally shocks Congress into rebuilding its own lawmaking capabilities, then perhaps some good will have come of it.", "url": "https://wpnews.pro/news/what-congress-actually-needs-to-do-about-ai-staff-up", "canonical_source": "https://prospect.org/2026/09/18/artificial-intelligence-congress-office-tchnology-assessment/", "published_at": "2026-09-18 09:30:00+00:00", "updated_at": "2026-09-18 09:53:27.295170+00:00", "lang": "en", "topics": ["ai-policy", "ai-safety", "artificial-intelligence"], "entities": ["Revolving Door Project", "Office of Technology Assessment", "Newt Gingrich", "Anthropic", "Jacob Coxon", "Dario Amodei", "OpenAI", "Sam Altman"], "alternates": {"html": "https://wpnews.pro/news/what-congress-actually-needs-to-do-about-ai-staff-up", "markdown": "https://wpnews.pro/news/what-congress-actually-needs-to-do-about-ai-staff-up.md", "text": "https://wpnews.pro/news/what-congress-actually-needs-to-do-about-ai-staff-up.txt", "jsonld": "https://wpnews.pro/news/what-congress-actually-needs-to-do-about-ai-staff-up.jsonld"}}