GeoLegal Weekly #12 - Robot Lobbyists and the Rule of Law
2024 lobbying season is underway in a big way. Why are lobbyists lawyers? Can AI automate lobbying? Does human and automated lobbying support or undermine the rule of law?
On the heels of defining Geolegal Risk last week, I’m now focused on fleshing out all the ways politics manifests through the legal vector to cause harm or opportunity. To do this, we need to think about all the ways law is created. One of the core ways this happens is through legislative policymaking and, one of the biggest influences on that is the lobbying process. The 70 elections underway in 2024 - especially the US Presidential and congressional elections - provide a live use case for understanding the geolegal risk related to lobbying and also how generative AI is going to change this.
By the end of this, you’ll not just be an expert on how lobbying is shaping up in the US election but also how generative AI can be harnessed to automate lobbying. We’ll also look at how AI-powered lobbying can undermine rule of law and think a bit about the evolving role of law firms in this dynamic.
To bring to life the post below, I spoke with Damien Riehl who is VP at VLEX and on the leadership team at SALI. There’s a lot in this interview - from broader discussions of AI and the law, but there’s a big emphasis on today’s topic, lobbying, where Damien scares me half to death by suggesting foundational AI models could be manipulated to throw in really dark suggestions when they receive lobbying requests and the like. You really need to watch it here.
2024 Means Game On
The 2024 election in the US is shaping up to be pretty close. This leads people to say that the election is a huge risk and a source of real uncertainty. The probability is tight right now but actually on some level this is the most forecastable election ever. That’s because it is an incumbent president running against a past president - in effect, we have more data than pretty much ever before to understand how they would govern.
That data is leading companies and foreign countries to position for influence. Japan, for example, has ramped up its lobbying in Washington, with one of their key lobbyists reiterating my point above:
“President Trump is a known quantity. You know what you're getting with him. If you're an ally of America, and you are America's friend, there's no better friend than President Trump. If you don't pay your bills, you're not America's ally when America needs you, I think you can understand, President Trump is not going to be that supportive of your agenda.”
Japan isn’t necessarily an outlier; there are many countries spending huge amounts of money in Washington. Data from OpenSecrets, which tracks all lobbying spending, shows the countries and entities spending the most in Washington (though there is a disclaimer that they are currently in the process of auditing their data, so there may be some errors).
The domestic agenda is also red hot. On the traditional side of the agenda, companies and taxpayer interest groups are jockeying for when a bevy of tax breaks expires at the end of next year. Banks, for instance, are employing more lobbyists than at any point in time before the financial crisis. Four of the biggest fund managers are paying millions to lobbyists to counter the anti-ESG movement (you can read my take on ESG politics here). A list of the largest lobbying spending in 2023 puts the US Chamber of Commerce, National Realtors Association and healthcare interests atop a list that quickly then is filled out by all the big tech, aerospace, and pharma companies. The oil industry’s top interest group is gearing up to spend eight-figures to “educate voters and policymakers on key energy policy issues ahead of the 2024 elections,” according to Fox News.
Of course, AI regulation and US-China relations also tops the list. Last year, 450 different companies and groups lobbied Congress on AI according to data collected by OpenSecrets as analyzed by the The New York Times. The same article reported this week that 100 tech leaders and top technology investors will meet in Washington in May to press Congress to do more to counter China and also to voice their opinions on how AI should be regulated. The Times reports:
Tech executives plan to use the event to directly lobby against A.I. regulations that they consider onerous, as well as ask for more government spending on the technology and research to support its development. They also plan to ask to relax immigration restrictions to bring more A.I. experts to the United States.
Crypto currency, which may have fallen off the front pages in recent months amid the AI hype, has advocates also pressing its case. The FT cites OpenSecrets data that the crypto lobby rose from $1.5mn in 2020 to $27mn in 2022. That’s nothing compared to the $80mn raised by crypto industry group “Fairshake”, which plans to deploy this money to target Senators in close races who have been crypto skeptics.
Why are lobbyists lawyers?
There is a large overlap in the Venn Diagram between lawyers/law firms and lobbyists. Many of the highest paid lobbying firms are law firms. Why is this the case? First of all, many members of Congress are lawyers and the most effective people at lobbying Congress are often former colleagues. Second, many lawyers evolve from helping clients navigate the government - whether defending them against government claims or producing regulatory analysis - to trying to influence it. Third, it’s been argued that the skillset of lawyers and lobbyists is very similar - for instance, a mastery of procedure and systematic thinking.
This is changing a bit. Increasingly, Congress is leaning on industry to provide data-backed analysis so the rise of Ph.d lobbyists is worth noting. But predominantly it remains a legal game. That the DC bar allows non-lawyers to own part of law firms subject to certain restrictions creates a pathway to partnership for non-lawyer lobbyists within law firms-further reinforcing those firms’ capabilities.
Lobbying and the Rule of Law
Before we get into how AI has the potential to shake up the lobbying industry, it’s worth pausing to situate lobbying in the broader constellation of rule of law: Does lobbying reinforce fundamental rights like the right to petition the government- or does it undermine representative democracy? There are a few arguments on both sides.
The positive spin on lobbying and the rule of law is that lobbying allows information transfer from industry experts to members of congress so they legislate better. Lobbying empowers citizens and interested parties to voice their views to the government, enhancing participation in democracy. Because lobbyists are meant to register - generating all that data I cited above - lobbying has the potential to increase transparency, particularly with respect to foreign and corporate interests. Lobbyists take on the drafting of much legislation which enables Congress to theoretically do more policymaking than if it had to produce all such legislation and analysis on its one.
On the flip side, lobbying has the potential to amplify the voices that are best funded and best networked. Lobbying can increase complexity of legislation by forcing “lowest common denominator” compromises - distorting the policy intention of a law. For as much as lobbyists are forced to register, unless C-Span starts live-streaming “the room where it happens”, conversations in private rooms of DC steakhouses are patently opaque. Lobbying can also redirect efforts of legislators to respond to their constituencies because they are instead worried about moneyed interests backing their challengers. Finally, even as Congress has become more diverse, lobbyists have not - I found this study particularly interesting.
While I’d argue that lobbying, in the best case scenarios, serves a fundamental role underpinning democracy it is no surprise that the general population feels lobbying and special interest groups have too much influence over Congress.
AI in Lobbying
The reason I’ve run through all of the above before getting to the juicy stuff is that we can’t measure the impact of AI on lobbying without acknowledging the state of lobbying today. Which is to say that many of the examples I will give below will immediately raise questions about undermining rule of law but that we have to acknowledge the current system is not exactly uncontroversial. Final point - I’m covering actual lobbying here; influence operations like deepfake voter manipulation are not the subject of this post.
So let’s get into the fun stuff. First of all, OpenAI took a position earlier this year it is: “still working to understand how effective our tools might be for personalized persuasion. Until we know more, we don’t allow people to build applications for political campaigning and lobbying.” That might be true with respect to apps but I’ve been able to recreate a number of the items below by asking simple prompts.
Two studies show how AI can create a new form of “astroturfing” - an act that effectively violates the principle of “one person one vote” by orchestrating an influence campaign from fake people or non-constituents posing as constituents for greater impact.
In the first study, Stanford Professor John Nay showed how scaled legislative analysis by an LLM can identify whether a proposed law is relevant or not to a corporation by digesting its public filings and, if relevant, generating a letter to a member of Congress suggesting changes. Nay raises a unique point that such AI-generated lobbying could distort the role that the law plays in providing information about citizen preferences, which would mis-train AI about citizen views in the future.
In the second, Sarah Kreps and Doug Kriner of Cornell University implemented a separate experiment at scale by sending over 30,000 letters to over 7,000 US state legislators in order to see if response rates varied by AI-generation vs human-generation. They start by noting that in 2017 when the Federal Communication Commission’s “net neutrality” rules were overwhelmed by public comments it was possible to flag repetitive or convoluted content. Their findings are particularly interesting:
“[W]e found that legislators were modestly less likely to respond to machine generated content than to human-written emails; however, although the difference in response rate was statistically significant, it was substantively small – less than 2%. On some topics, there were no differences in response rates between the human and machine-generated emails; and for one topic machine-generated correspondences elicited a higher response rate than human emails, although this difference did not reach statistical significance. Further, a sizable number of AI-written correspondences elicited lengthy and personal responses suggesting that legislators believed that they were responding to constituents. While not a perfect predictor of legislative action, responsiveness is a valuable proxy for legislative priorities given the demands on legislators’ time (Costa 2017; Butler and Broockman 2011; Einstein and Glick 2017), and the decision to answer a constituent letter implies a calculus about the importance of responsiveness to that individual or issue (Bol et al 2020)”
Nathan E. Sanders and Bruce Schneier take this a step further in MIT Technology Review by explaining how AI can be used to build a “micro-legislator” or a lobbying robot that digests proposed legislation, understands the smallest, least-detectable changes to legislation that could benefit a particular interest and then strategizes how to get that language inserted by pulling levers of power. They write:
Put together, these three components could create an automatic system for generating profitable microlegislation. The policy proposal system would create millions, even billions, of possible amendments. The impact assessor would identify the few that promise to be most profitable to the client. And the lobbying strategy tool would produce a blueprint for getting them passed.
One implication I worry about is that as there is more noise in the commenting process, the process could be thrown away (though improved detection mechanisms could limit this). This scenario would undermine the lawmaking process if it is not replaced by alternative forms of citizens petitioning the government. As Bridget C.E. Dooling and Mark Febrizio highlight in a Brookings paper:
By reducing the costs of producing “malattributed” comments, generative AI could lead to a pooling equilibrium—to borrow a concept from game theory that is often applied to insurance markets—where agencies can no longer meaningfully distinguish between valid and malicious comments. Agencies could then be inclined to assume all comments might be “fake” and discount their relevance, weakening public commenting as an avenue for meaningful public input and the formulation of improved policies and, ultimately, making people worse off.
What happens if regulatory commentary and direct letter writing to Congress is discounted? I suspect we’d see members of Congress fall back on leaning on human interaction. Those interactions can take two forms. First, it might imply an elevated primacy of those “smoke-filled room” interactions where human lobbyists who are already known to members reaffirm their advantage in influencing policy. In the case of the micro-legislator, Sanders and Schneir note that this would advantage those already with power and connections. Second, this would create incentives for citizens to use other means of expressing their policy views such as protest. Movements like Occupy Wall Street are an example of the types of ways citizens might express themselves if other avenues to government are shut down.
From here to there
No doubt the above presents both ethical and practical challenges for lobbyists and corporate government relations teams, who want to leverage technology but wouldn’t want to be seen as using technology unethically. A Bloomberg Government analysis showed a number of lobbyists hand-wringing about this. I think, however, that its only a matter of time before the approaches above are refined and mainstreamed, making it seem natural to use them to amplify impact.
In fact, while some of it seems hypothetical, there are a number of tools in the modern influence toolkit that already are pretty powerful. FiscalNote has built a comprehensive advocacy stack that brings together coalition building and policy analysis alongside media outlets they have acquired, which create further opportunities through advertising. They also have a platform for corporate public policy teams to monitor regulation and coordinate action. Abstract, a relatively new start-up, similarly monitors laws and regulations but has a more focused emphasis on risk analysis using AI, “abstracting for what is important.” Another company, AXIS, takes the analysis dimension one step further by developing network maps of the powerful across industries and sectors globally. Their demo below is pretty cool.
I hope you’ve found the above a unique overview - starting with the political state of play, running through some theory and then looking at where the tech is today and where it could go. If you found it useful, tell a friend.
-SW