Civil Rights, the Notice Regime, and Illinois' 2026 AI Hiring Law with Jason Rosensweig

Khullani Abdullahi (00:01.7) Good afternoon and welcome to the AI in Chicago podcast. I'm your host, Khullani Abdullahi, the founder of Techné AI, a Chicago-based AI governance advisory firm. AI in Chicago spotlights the operators, builders, and thinkers scaling applied AI from the heart of the Midwest. Each episode delivers practical stories and playbooks leaders can use to win with AI minus all of the hype. When Illinois decided to regulate AI in hiring, it made a choice most states didn't. It didn't pass a standalone AI statute. It didn't mandate bias audits. It didn't adopt the Colorado consumer-protection frame or the New York City audit model. Instead, Illinois put AI squarely inside the Illinois Human Rights Act, treating algorithmic discrimination as a civil rights problem, enforced by the same agency that handles housing, credit, real estate, and public accommodations complaints. That choice has implications most of the national commentary has missed. The disparate impact standard means intent isn't relevant — only effect matters. Zip code is mentioned explicitly as a prohibited proxy. Vendors aren't directly regulated, but employers remain liable for the tools they buy. Today, I'm so pleased to welcome Jason Rosensweig, who serves as Senior Policy Advisor at the Illinois Department of Human Rights. Jason, welcome. Jason Rosensweig (01:26.9) Thank you so much for having me, Khullani. Khullani Abdullahi (01:29.6) Of course. Jason helps shape the state's approach to civil rights enforcement in an era when hiring, housing, and credit decisions are increasingly made by algorithms. Before stepping into the advisory role in January of 2025, he served more than two years as IDHR's Director of Legislative Affairs and Policy, and he continues to serve as a commissioner on the Illinois Commission on Discrimination and Hate Crimes, a role he has held since 2021. Jason brings an unusual intellectual pedigree to civil rights policy. He holds a PhD from the University of Chicago's Committee on Social Thought, where his dissertation, "Progress, Forms of Life, and the Nature of the Political," examined community pluralism and the tension between idealism and realism in democratic life. He earned his BA in comparative literature and an MA in French literature from Stanford and was a visiting scholar in the Faculty of Philosophy at Cambridge. Outside of IDHR, Jason teaches Freedom, Equality, and Justice in American Government as an adjunct professor with the Northwestern Prison Education Project. He previously led the Shalom Hartman Institute's Chicago office and served as assistant director of advocacy and international affairs at the American Jewish Committee. He was a member of the Chicago Council on Global Affairs Emerging Leaders Class of 2021. Jason, how did you go from literature to leading policy in state government at the Illinois Department of Human Rights — and also teaching in prison? You've moved from academia to civic leadership to government. How did you end up here? Tell us your origin story. Jason Rosensweig (03:12.3) It took a few steps, but it makes sense to me. I grew up in the DC area. Both of my parents did federal government work — my dad did international aid work and my mom did program evaluation. Government and policy was always a strong part of things. But when I went to college, I fell in love with the humanities, and academia in the humanities felt like the vocation for a long time. I drifted from literature first to philosophy, partly because I felt like maybe I could solve something a little more in philosophy — which turned out not to be right. And when I was choosing my dissertation, I chose political philosophy in part because it's grounded more than other parts of philosophy; it has to be tethered to reality and history in a different way. My focus was on what it takes for us to get along in a free and pluralist society — what we need to share for that to work. And the answer was actually kind of a lot. It's not just having the right constitution. The concept was a form of life: you have to share a form of life, and a fair amount of trust and expectations and things that aren't always visible, to make community work. Part of that was a belief in localism and the importance of the state and local side of things. So I moved to DC briefly after grad school here in Chicago, but then made a choice to come back here to do policy and public service in a place rather than at the national or international level. First I worked in the Jewish world at two different organizations and learned the advocacy game and playbook, and came to understand pluralism from the perspective of a particular community. Then I started to get involved in more state policy work on hate crime and anti-Semitism, and got appointed to the Commission on Hate Crimes, which is part of the Department of Human Rights and chaired by the director. That's what led to my joining the state. I like being where the task is to be responsible for everyone and not just a particular group. Khullani Abdullahi (06:07.8) It's interesting, this arc of your career. A question from your dissertation: you mentioned that the right constitutional framework — the legal documents — are not sufficient to build the kind of society where people trust each other, get along, and can govern themselves well to an end that benefits most, if not all, of its citizens. What were some of the key necessary ingredients you discovered in your research for building that kind of society? I ask because this political macro-environment moment we live in feels like an opportunity to ask that question and then start to build, and maybe do some rebuilding. Jason Rosensweig (07:11.1) I think we've certainly seen in the last few years that the constitution is, in many ways, vulnerable — it depends on people to see it and use it in the right way. Especially if you have a president who's willing to apply sufficient force to push against it in different ways, it's not enough just to have the right words on the page. Khullani Abdullahi (07:45.2) So in your research, are there factors we can think about for rejuvenating American society that would help support the vulnerability of our constitution — especially because at all levels of government we're seeing poor actors? And by poor I mean actors who lack integrity, who don't believe in the truth, who are willing to lie, cheat, and steal and ignore the norms and laws. It turns out that if you don't have decent people in these roles, the entire edifice kind of collapses. So is it the right people in the right place? Are we just not raising moral and rational Americans? Where have we failed? Jason Rosensweig (08:48.5) There's no one answer, but one of the concepts I think about a lot in this area is trust. You need basic default social trust, but you also need trust in institutions, in government, in education and schooling. And that's all been significantly degraded. I absolutely think that rebuilding those different aspects of trust would go a long way. Obviously it's not easy. I always think about Exodus in the Bible, or Plato's Republic, where the idea was that the only way to create the ideal society is to wait until everyone in the current society is gone. Khullani Abdullahi (09:46.6) In the Republic, I believe it was to remove everyone over the age of ten so they don't inherit the memories and failures of their parents. The trust element — this is unrelated to the AI in Chicago podcast; it was my own curiosity when you mentioned the elements of your thesis — but it's very timely for the political moment we're in. Jason Rosensweig (10:22.0) I could tie it to the AI and tech policy issues. Khullani Abdullahi (10:25.2) Yes, let's do it. Jason Rosensweig (10:29.8) Because the notion of what a proper society is, what social health is, what the right patterns of behavior around social interaction and relationships are — technology and social media are problematic in many ways for all of that. They're obviously helpful in many ways too, but it certainly feels like we're trying to find a new balance. Some of that is a policy and law question, but some of it is an individual-choice question — between lived experience with people in "meat space," as some people say, versus just the digital. Khullani Abdullahi (11:21.9) I don't know if you've seen this, but there's a trend I think is accelerating, and it makes my liberal arts heart very happy: Google DeepMind just brought on a philosopher. A number of philosophers are at Anthropic. We're seeing them at OpenAI. It feels like a moment for the humanities — and in particular, philosophers and philosophical thinking are ascendant. And it's precisely for the reasons you articulated: we have to ask what it means to live well, what it means to live well together, what it means to become the kind of people who can live well during a moment of such transformation. Those are precisely the kinds of questions that discipline is designed to address. There are people whose political rights have been deprived because they've been convicted of a crime, and I think it's fascinating to take political philosophy and teach it to incarcerated students. I'd love for you to talk through what that looks like, and how they experience their own incarceration differently in light of the work you're teaching them. Tell us about the Northwestern Prison Education Project and what that work looks like. Jason Rosensweig (13:25.2) Your question gets at one of the most notable parts of the experience. In contrast to the students I taught at the University of Chicago or American University, you can't take anything for granted about them having a basic sense of trust or respect for systems, laws, governments, and the path of history — everything has to be earned in a way. That was very potently challenging, in a good way. The next time I teach, I'm hoping I can teach some more pure philosophy, because they really responded well when we did that and avoided some of the charged historical parts of the American story. But overall, I can't talk about this without saying the students are amazing. It's selective — a lot apply and only some are able to do the program — but the people who had been there for 30 years, since they were 18, and had every right not to have aspirations, had somehow transmuted that experience into real wisdom and a desire to do good. I went to the final session of another class on public speaking, where each student gave a five-minute speech about a project or proposal for what they wanted to do when they got out. They were amazing — someone who had a young baby when he was first incarcerated had been thinking about what it's like to be a parent with a little kid. They challenge me because they don't take anything for granted, they care, and they're not just there to get through it, which can happen with other students. Khullani Abdullahi (15:51.6) The material feels much more consequential in and of itself — it's not for a grade or a diploma. There's this arc of moral operationalization that's a common thread throughout your work: a continuous attempt to formalize and operationalize these key values. And now you're inside a state agency. What I've discovered, for myself and my audience, is that most of us don't know how agencies work day to day. So what's a misunderstood thing about civil rights enforcement at IDHR from a day-to-day standpoint, and how should people think about what IDHR does? Jason Rosensweig (17:08.0) The most common misconception we have to deal with is the notion that if you're the Department of Human Rights and something unjust is happening to me, you should be able to do something about it. And yes — but within certain parameters. There needs to be discrimination based on a protected class, and there needs to be some minimal bit of evidence to that effect at the first stage. It's very hard when people have clearly been treated unfairly or wrongly, even significantly, but it might just not fall under our jurisdiction or what we can do. Khullani Abdullahi (18:06.2) So the first issue is people understanding the kinds of harms they can even bring to you, or that you could potentially remedy. In Illinois, IDHR is uniquely designated — I think uniquely in the country — as the rulemaking and enforcement body for the use of AI in employment. From a policy-design standpoint, why did that architectural choice matter? Jason Rosensweig (18:48.9) It was an intentional choice. The legislature — the Senate, at least — made a decision, at least initially (it might be different this year, we'll see), that they wanted to amend existing statutes rather than pass an omnibus AI regulation bill. And there's a particular origin story to the bill. Representative Jamie Andrade initially proposed it. This gets at what you noted — the presence of zip code as a proxy for race, which seems a little out of place. The original bill was just focused on using zip codes as a proxy for race. We said, if we're going to do this, it needs to be a little broader than that. But that stayed in as a remnant of the original version. It wasn't our initiative, but we did help redraft it and create the notice concept and regime as the most effective and reasonable way to go about it. I think it's likely, or at least possible, that there will be other bills passed this year that address other aspects of employment in statutes outside the Human Rights Act. Khullani Abdullahi (20:21.2) There are dozens of AI-related bills in committee in this 104th General Assembly. We'll see what shakes out by August when the governor is signing bills. So, thinking about the doctrine and the origin of the AI-in-employment bill that IDHR is responsible for regulating: what was it about algorithmic decision-making that was the crux of the issue? And then I also want to talk about the stakeholders you've heard from. Jason Rosensweig (21:14.6) Uses of technology have always carried liability. You can't say, "I fired that person, but I only did it because the computer told me to" — it's always been the case that there's still liability there. So part of this bill was just saying very clearly and explicitly, in bright letters, that that's the case. But there's also a new issue with the new AI tools, particularly the LLMs: there's insufficient visibility into what's inside and why it's doing what it's doing. That creates a special gap that the notice is designed to address — to say, no, you need to be able to say some things and have looked at what's inside, how it works, and what it's doing. Khullani Abdullahi (22:13.6) One thing that's interesting about the legislation — and I think it's unique in the draft rules — is the "influence or facilitate" language in Subpart J. I take that to mean, and I'd like to confirm with you, that it's not even that the algorithms make hiring or firing decisions that are then found to have disparate impact on protected groups. It's that if they even influence, shape, or create the conditions — like an automated queue — then the final decision doesn't have to be driven by an automated system. If it's created conditions that give you a biased pool, that's sufficient for employers to be concerned. Jason Rosensweig (23:24.8) Here's an example that might help. Imagine you're hiring for a position and you use a tool to score a resume or a video interview, and it's a low score. That's not the decision — that's just a score that a human then looks at. But if that tool, let's say, can't read the face of someone with a disability in the right way in the video-interview analysis, that's discriminatory. Khullani Abdullahi (24:04.8) And even if that was only one element of the decision-making process you pulled together. That's helpful, thank you. I'm interested in the stakeholder conversations, because I suspect you got very different viewpoints depending on whether you were talking to employers or unions. You shared earlier that you did a significant, in-depth, multi-stakeholder engagement process. I'd love to hear how it was structured, what those conversations looked like, and how you engaged the different stakeholders in developing the rules. Walk us through it — most people don't have a clear understanding of how a rules-based order is developed around a bill. The legislature hands you things, and then the state agencies have to figure out how to operationalize and bring them to life. Jason Rosensweig (28:40.0) It started with research and consultation with subject-matter experts of different sorts — lawyers, technologists, people who'd come from the White House Office of Science and Technology Policy — to get a first lay of the land, and also consulting other states like New York that had done something along these lines. But it quickly became clear there was no standard answer for how this would work; it was so sector-specific and perspective-specific. A union is going to have a very different perspective from a hospital, even in the same sector. I think we had maybe 15 bilateral meetings and then some larger group conversations. A perfect example is the nurses. They were saying, "We're seeing that hospitals are using AI to look at the electronic health records of patients — this person had a drop in pulse-ox and it took this many minutes for the nurse to arrive — and they use that to create assessments and evaluations of nurse performance. But they never told us about it. They never asked us. We don't know how it works." So that's a good example of a very specific use of AI in a sector that we wouldn't necessarily have thought of. One of the initial impetuses for the broader engagement was thinking about the different kinds of notice — is it a website posting? A letter? A physical posting at the workplace? — so we needed to talk to a lot of different businesses and sectors to see what would be effective as broadly as possible. Khullani Abdullahi (31:00.4) Especially because a lot of companies don't even have a physical workspace anymore. It used to be that all those labor-related notices would go wherever the coffee and water cooler were, and then you were covered, because everyone at some point was going to grab coffee or water. Jason Rosensweig (31:19.0) Which still works in a lot of cases, but not all. Khullani Abdullahi (31:30.2) Even talking to different stakeholder groups within a specific industry revealed all these numerous use cases of AI in the workplace — beyond just the decision to hire or fire, but also evaluation, promotion, retention, and performance improvement plans. So if you could bucket the stakeholders: you mentioned unions and healthcare provider organizations. Did you talk to construction? Utilities? Financial services firms? What was the industry coverage? Jason Rosensweig (32:17.0) It can't cover everything, but we covered some of the larger tech industry groups and some of the larger business and employer groups — retail manufacturers, the restaurant association, the chamber of commerce, those broad employer and industry groups — and then a fair number of associations. We talked to a nurses' union, but we also talked to a nursing association. Staffing firms were another one, because they're a kind of intermediate employer where particular things come up. Khullani Abdullahi (32:57.3) What were the top employer pushbacks? Jason Rosensweig (33:14.2) We definitely heard more than once, "Couldn't you put this burden on the developers instead of the employers?" And — yeah, but that's not what the legislature did here, and there are reasons why. Employers need to have some responsibility. There's room maybe to have some responsibility on the developers too. A second thing: some folks would have enjoyed just one straightforward requirement — okay, it's written, you have to mail it or email it, one simple thing. And there was a little pushback about having to say what we're doing — whether you use trade-secrets language or other proprietary concerns, or just "we don't want to have to do that, and we don't think we should have to." Khullani Abdullahi (34:28.8) What I've found, as someone who helps companies bring their responsible-AI and AI governance, risk, and management initiatives to life, is that they underestimate — and we as a society underestimate — the amount of work required to actually do that. Even if you have an obligation as an employer to inform your employees and prospective employees about how you use AI, there's a big lift in ensuring your team and organization is even literate about those obligations and capable of carrying them out. And the legal layer and the executive layer aren't usually where the violations happen — it's downstream, where the hiring managers, HR, and the engineering team acquiring these vendors and software subscriptions are. That's where the rubber meets the road for compliance. When you were engaging stakeholders and promulgating the rules, did anyone feel they'd be ill-equipped to ensure their teams could meet these mandates? Jason Rosensweig (36:08.2) The main place that comes up is with small businesses, which can be one person. I'm very sensitive to that concern, because it's a lot to ask every small-business owner to understand all of this. The legislation didn't permit us to have different standards for small businesses. One of my hopes is that various sources will play a role in helping them. Some of that is us — issuing guidance and templates. I think there's a role for trade associations to say, "For our members, we're going to do an analysis of the five main platforms you're all looking at for X, and give you some information about what you'd be getting." And obviously the vendors can help too, because what's being asked in this notice is basically: here's the product we use to do this thing for this position. Vendors will be able to help with that. Khullani Abdullahi (37:34.1) So when people think of IDHR, it's primarily employment, and we've talked mostly about the rules you're promulgating on AI in employment. But you also have jurisdiction over housing, financial credit, real estate, and public accommodations. Are you seeing AI show up across those domains already, independent of any bills or rules? Is the public in Illinois coming to you with complaints in those sectors? Jason Rosensweig (38:18.7) We're certainly seeing uses — and violations — of AI and technology in those areas. There was a federal settlement, I think with DOJ, with an Illinois bank that had been doing a version of redlining; the settlement was canceled this year. But we're certainly seeing those, especially in housing — tenant screening, and some of the lending. When it comes to actual cases that come to the department, that's part of the reason we need the notice: you're not going to know that you got fired, or whatever, because of the technology they used unless you know they were using it. So there's a bit of a chicken-and-egg question there too. Khullani Abdullahi (39:16.7) The notice provides people with the impetus to reach out and inform you of their treatment. One of the things executives struggle with is staying up to date. Here you are, a senior policy advisor with a broad portfolio, and AI is now upending every element of that portfolio. How are you, as Jason, keeping up to date? What are you reading and listening to? How much of your time goes to staying current? Because it's probably the principal complaint I hear from executives — how do we even stay abreast of all these changes while still doing our full-time jobs and having families? What's working for you that we could steal? Jason Rosensweig (40:22.2) There's no doubt it could easily be a full-time job to stay abreast — a full-time job I'd enjoy, but it's not the only thing I have to do. So I try to identify particular steady sources of information, but also to reach out to people. I reached out to someone I saw testifying at the Senate hearing last week who seemed interesting to learn from. And I'm a particular fan — happy to plug Casey Newton's Platformer. It's a really excellent and consistent source of insight and information of all kinds. Khullani Abdullahi (41:12.7) Do you read a lot of Substacks? Jason Rosensweig (41:17.3) Yeah, there's a steady diet. That's one of the ones I find most compelling and have subscribed to for a long time, but there are a few others as well. Khullani Abdullahi (41:32.4) My podcast initiatives aren't that sophisticated yet, but sometime this year I want to take those links and readings and add them to the show notes. Since this is a hobby, we haven't gotten to that point — and please, all the podcast vendors, don't email me; they're always telling me they can help make my podcast more professional, and I'm not quite ready for that. So one question: I think it's a little unclear in the content of the bills — the anti-discrimination principle doesn't seem to require AI-specific statutory language to reach algorithmic discrimination in housing or credit. So do you, and does the organization, believe the anti-discrimination principles apply even beyond employment, even though only the employment provisions were amended? Jason Rosensweig (42:47.4) As I was saying before, we certainly take the position — and are confident in it — that the core discrimination protections apply: the fact that technology was the reason, or was what you were using, is not an excuse if the result is unlawful discrimination. But we think there's a lot of value in saying, in bright letters, explicitly, that that's the case, so people don't have to intuit or infer it. The biggest meat in the bill, though, is the notice. That's what will really be new. Khullani Abdullahi (43:37.5) It's interesting, because I wonder — there will be a certain class of plaintiffs' firms and civil rights organizations who think about this as an expansive umbrella, a class of new rights almost. With traditional disparate impact, you now have the AI-vendor angle, where you don't really have to point to a specific person if the corporation is using a specific vendor, and then you apply the disparate-impact standard, which doesn't require intent. I do think — and this is probably one of the claims businesses will make — that increased cases and litigation will help refine both the rules and the statutes. That's something unique to the American legal context: it's iterative. You have legislative intent, you have administrative rules, and then you have what happens in court in front of a jury and a judge, and all of that is tied together. In your role, with this transformative technology coming like a freight train and being embedded into all sectors of society, you have a delayed ability to wait for a bill, promulgate rules, and then enforce them. The common complaint lobbed against government and state agencies is that they're not moving fast enough — that this is a technology evolving at a pace we've never seen before. Some of that is true; these feel like unprecedented times. So how are you at IDHR adapting to this accelerated environment, with new kinds of claims and complaints arising and new legislation being handed to you to build rules for the first time? Is that pace being challenged, and are you making adjustments to deal with the deluge, or is it business as usual? Jason Rosensweig (48:04.1) A fair amount of the real effects and changes we're going to have to tangle with have yet to fully materialize. But because it's so slow to develop the processes — to hire staff with new expertise in AI or tech — there's a need to get ahead a bit. One way to think about the notice requirement is that it's about surfacing more information about what's going on in order to make better policy and laws going forward. In terms of adopting AI, the State of Illinois has broadly been cautious, and I think that's appropriate — states should be more cautious than the private sector. There's a different kind of responsibility, and there should be less tolerance for error and failure. The argument that this is moving so fast that we can't do anything about it is, to me, not always made in good faith. And going back to the beginning of our conversation about the role of a philosopher, or civil rights, in this — I don't think civil rights people or philosophers should be deciding all of this, but it's an important voice, because if you're focused just on how to make the thing faster or better, you're going to miss some of the broader effects and consequences. Khullani Abdullahi (50:01.1) Absolutely. And alongside that, the bad-faith arguments want government to be paralyzed. In the testimony I submitted before the Senate subcommittee on social media and AI, I called out that industry would also likely argue we should just wait for federal legislation — knowing there's no path to federal legislation, which is another way of saying do nothing. But to be fair, there is an argument, and I find it compelling, that it needs to be iterative. So I'd love to hear where updates might come from. A concrete example: in the EU AI Act, in SB 53, in the New York RAISE Act, and even in Illinois' frontier AI bill, we have these compute thresholds for training — something like 10 to the 26th FLOPs, don't hold me to that. Those thresholds are anchored around the compute required to train a frontier model two or three years ago. That's not the case in 2026. Chinese open-source model companies have already shown those thresholds have shifted, and increases in algorithmic efficiency, new computing substrates, and new training methodologies mean 10-to-the-26 FLOPs is no longer some immutable proxy for the capacity to train a frontier model. It took years — the EU Act in particular — to develop that standard, but the rate of improvement means those thresholds get out of date quickly. So, one: are you and other state agencies thinking about that at IDHR? And what room is there, outside of redrafting legislation, to adjust those thresholds? Do you anticipate the rules evolving, or are they promulgated and then we live with them until another amendment or a new bill? Jason Rosensweig (53:08.3) We tried to design our notice to be as flexible as possible for that reason. But I'd certainly think it's possible, and probably likely, that at some point — whether it's two years or ten years out — either the legislation or the rules would need to be amended. Section 230 might be a lesson here: that was put in well before most of the things it applies to now existed, and it was never revisited, and probably should have been. Another concept out there — and I'm not sure it's the right rule of thumb, but it's interesting — is that you focus more on regulating the uses rather than the technology itself, which the AI notice in employment is a version of. Khullani Abdullahi (54:17.0) Then you have more longevity, because you're regulating and prohibiting certain kinds of behavior and outcomes rather than the mechanisms by which they're facilitated. That's a good point. It's always important to understand both sides of that argument. I realize a lot of our conversation assumes familiarity with the bills being considered, so I'll add an element at the beginning of the episode to explain what IDHR is promulgating rules around so people don't lose the thread. But I want to give you an opportunity to speak directly to the vendors — because they're not specifically called out, but any employer worth their salt is going to have vendors indemnify them. There's no way I'm not going to shift the balance of that risk and liability to the extent I can. So those vendor contracts will have to be updated — indemnification clauses, insurance — there are nitty-gritty, tactical, on-the-ground changes coming in light of Illinois's slew of AI bills, including the Wellness and Oversight for Psychological Resources Act from August 2025. So what do you want vendors — the companies actually building and deploying these systems and making them available in Illinois — to know? Illinois has the third-largest city in America, a $1.2 trillion GDP, 12 million people. They're not going to leave the market; they're going to adapt. So what do you want them to take away from this law? Jason Rosensweig (56:35.8) We really want vendors to think, as they're building products, selling them, and negotiating contracts, that there is some accountability for safety and responsibility down the road. That's the biggest thing. And this connects to disparate impact, which you mentioned — not everyone may understand it. Most discrimination law is about disparate treatment, which is like: a man and a woman apply for a job, the man gets it because he's a man and the woman doesn't because she's a woman. That's the discrimination. Disparate impact is more like redlining: "No, we're not going to lend to the South Side of Chicago, but it's not because of race — it's just because of zip codes." Disparate impact is a particular form of liability that's often going to be where the problem in AI is, because it's not going to be "I'm going to go in and warp the results to harm Khullani" — it's just going to do that on its own. And it's rational for vendors and developers not to prioritize that. Let's say half a percent of the population is going to be outright eliminated off the bat, unfairly. If you're hiring for a position, you might think, "I'm still getting 99.5% of the applicant pool — that's enough for me to find the people I need." Khullani Abdullahi (58:23.5) Unless you're in that small group of individuals who are routinely excluded. It's interesting how fast this has emerged. In 2016, the first articles about facial-recognition software and its inability to capture melanated skin correctly were coming out. At the time, people hadn't been arrested erroneously. But in the decade since I authored that article in 2016, we've seen numerous times that AI systems don't work well for melanated skin. It's also important to highlight that we've seen white people arrested erroneously due to these same systems. So the problem isn't just disparate impact for some groups — there are fundamental limitations where a portion of the population will experience disparate impact even if they belong to the majority race the data was trained on. There was recently a white woman in the news, and a white man in the South. So it's been fascinating to see vendors continue to move in the wrong direction when it comes to the surveillance implications of these technologies. Jason, thank you for sharing your insights. I usually like to wrap up with some final thoughts and then a lightning round on your favorite things in the city — favorite thing to do in the summer, thin crust or deep dish and favorite pizza place, favorite coffee house or bar. But first, some final thoughts on the work you do, the role you and IDHR play, and what people can expect to partner with IDHR on this year and beyond. Jason Rosensweig (01:00:50.2) We always want to hear from people about what problems they're seeing — where discrimination and injustice are happening and redress isn't available for some reason. There's a whole other side of my job about adapting to federal changes in the last year, and without getting too deep into the specifics, the result is that states have to step up and take on a stronger, more substantial role in civil rights law and enforcement. That's probably here to stay, and it could be good for some of the localism reasons I mentioned before. One of the reasons I like this job and working for the agency is that it gives me a way to think about and help with these broader questions about social health — which we don't really have agencies meant for. Some other countries have a ministry of the interior or a home ministry that might get into that; here, public health has, in recent years, been one of the places where some of that broader social-health thinking has happened. Given the fraying we're seeing in our social fabric, it's particularly important to have a strong voice in all of that. And lastly — some people might plug a book — but on disparate impact, the legislation we have this year, SB 3777, the Civil Rights Safeguard Act, is explicitly meant to shore up state disparate-impact protections, because of an executive order seeking to eliminate them. Khullani Abdullahi (01:03:03.5) And because of a Supreme Court and a fractured circuit-court environment, the courts and the executive branch are leaving it to states to fill that role. Those are really great points to wrap your insights on. You're not a native Chicagoan — you're a native of DC — but Chicago is home now and has been for, I think, decades. You did a lot of your education here, and your immediate family is here. Tell us your favorite things about Chicago — food, pizza, favorite things to do in the summer. Where can people find you out and about? Jason Rosensweig (01:04:05.7) Especially as my social calendar and extracurriculars have gotten confined with little kids, one of the things I still prioritize is live music. I try to go to a couple of shows a month, and I think Chicago has a fantastic live-music scene. Khullani Abdullahi (01:04:22.4) A couple of shows a month with two kids under five? Jason, you're living a great life. Jason Rosensweig (01:04:28.0) Sometimes I sneak out after they go to bed and miss the opener. Khullani Abdullahi (01:04:32.0) You make it work. Where do you go for live shows? Any venues you really like? Jason Rosensweig (01:04:42.3) Chicago has a fantastic live-music scene. As a large venue, the Salt Shed has been a great addition. I really like Thalia Hall, and SPACE in Evanston. And this new place in the West Loop, Garcia's, is really special — small and intimate, but they get great acts. It's been my favorite new place in the last year. Khullani Abdullahi (01:05:07.0) I'm going to write that down — Garcia's in the West Loop. I do a terrible job of getting out to enjoy Chicago, so this question is really for me. Are you deep dish or thin crust? Does DC have a pizza it's known for? Jason Rosensweig (01:05:27.6) They have pizza, but I'm not aware of any particular DC pizza. As for deep dish, I'm like most Chicagoans — deep dish isn't the pizza we usually eat. For classic Chicago places, I think Pequod's is awesome, and I really like Paulie Gee's. And it's not Chicago style, it's Neapolitan, but Spacca Napoli is really good. Khullani Abdullahi (01:05:55.3) Excellent. Well, Jason, thank you for taking the time to share your origin story — this is a clear example of the importance of the humanities. It's almost applied humanities: you took all of those lessons and that research and now you're actively in government trying to improve society. So thank you for the work you do, and thank you for being so candid in your comments. Wishing you all the luck as you promulgate these rules — I'll be sure to weigh in with my public comments, because I have some strong opinions. And I'm looking forward to putting together some playbooks for the small businesses you mentioned, because there are ways for companies like mine to help streamline compliance so it's not costly or burdensome. Jason Rosensweig (01:06:50.0) Perhaps we could talk about that when we get to some of the other guidance and information we put out down the line. Khullani Abdullahi (01:06:57.8) Absolutely. Have a great day, and thank you. Jason Rosensweig (01:07:01.5) Thanks so much for the opportunity. I really enjoyed it.