In the new era of global competition, being at the forefront of Super Intelligence innovation equals geopolitical power. To maintain leadership and establish the next world order, the U.S. is moving to cement its lead.
Barnett Koven is a Managing Director of AI and Data at Accenture Federal Services. He came to Accenture Federal Services after stints in industry and a faculty job at the University of Maryland, where his counterinsurgency research showed that some of the most influential quantitative studies got the answer exactly backwards.
We discuss…
Why “inaction is in and of itself risky” — and how government risk-aversion compounds against less-cautious competitors
The “gun-armor race” in classified AI environments, where defense costs grow exponentially against linear offense
Entity resolution and predictive modeling for BIS export-control enforcement, and unwrapping the PLA’s shell-company maze
DOE’s Genesis Mission and Accenture Federal’s “backbone for scientific research”
Thank you to Accenture Federal Services for bringing us this episode.
Listen now on your favorite podcast app.
From COIN to Genesis
Jordan Schneider: Let’s do a bit of the career arc. How do we get from a PhD on COIN to DOE and the Genesis mission?
Barnett Koven: Absolutely. I spent a lot of time in Latin America and saw the aftereffects of some virulent insurgencies in Peru and Colombia in particular, ended up doing a PhD studying the spatiotemporal diffusion of insurgent violence as a result of counterinsurgency strategy — specifically as a result of development assistance and other non-kinetic interventions during counterinsurgency. Really loved that, and sought out the opportunity to do more policy-relevant research, which is not necessarily the norm at the academy. Ended up in a faculty job at the University of Maryland, which was also wonderful — a research institution where I was encouraged and able to do a bit more government-funded research as well as the traditional teaching and publishing you'd expect from a professor. Enjoyed that immensely.
However, I did a lot of mixed-methods work, and it became harder and harder to do the travel that I wanted to do and the data collection. At that point I had a number of friends at a competitive consulting firm who had reached out to me over the years and tried to get me to come join them. I finally said, you know what, what the heck, let me at least have the conversation. And I realized through some really influential leaders at that firm that this would be an opportunity to apply the rigorous research methodology and quantitative skills I learned in grad school and honed at UMD to real-world problems, and do it in a way that felt closer to the mission and closer to the policymakers and government who were actually hopefully benefiting from and using those insights.
Jordan Schneider: Let’s talk through some of those. What were some of the things that came your way?
Barnett Koven: In general terms, two of the coolest projects were using natural language processing in order to extract insights from diplomatic cables. You may not know this, but the Department of State sends back on the order of 3,000 cables between posts and headquarters in DC every day. These are chock-full of insights in textual form, but no one's reading 3,000 cables, let alone memorizing them and absorbing all of that on a daily basis. So the goal here was to use old-school, pre-gen-AI-revolution natural language processing approaches to extract and categorize some of that insight to make it easily searchable so that it could inform policymaking.
And I'm very, very proud of the work we had done there. That being said, if you fast-forward half a decade to today, I would have done it completely differently. Some of the tools that are now available that weren't then would have accelerated what was a six-month fast-paced project involving a half-dozen-person team into something that two people could have done in a matter of weeks, and just as well or better.
True Believers and the Risk-Averse
Jordan Schneider: Well, let's talk about that for a second. The idea that software is really cheap, anyone can run Claude Code, Codex stuff — in your interactions with the government, to what extent has the light bulb gone off that just a lot more is possible than even was, I don't know, 12 months ago?
Barnett Koven: The cop-out answer here is: it depends. And certain government customers are way closer to the forefront than others. Part of this is a product of their unique mission and how applicable it is to these changes. But part of it is also just risk aversion. Some of our law enforcement clients, for example, tend to exhibit a little bit more risk aversion here. That being said, I think some of the skepticism is really healthy.
Jordan Schneider: Wait, sorry — as in, I can't prove a case if an LLM collects the evidence for me? What do you mean by that?
Barnett Koven: I just meant nervous about second- and third-order implications, whether that's sensitive information being exposed to the public, or the evidentiary value being diminished because of a court ruling — and there have been court rulings around gen-AI-produced legal briefs and what have you. So I didn't mean anything specific, but there's a number of different directions to take that.
Jordan Schneider: Gotcha. Okay, but the ones who are the believers — what's the common line there?
Barnett Koven: There are a couple of explanations. Some of the true believers are coming from industry into government for the first time and have seen the power of these tools in their prior jobs in corporate America, where the uptick has maybe occurred first. Others are just super passionate about their mission, and while they've been in government for perhaps decades in some cases, they make it a point to stay abreast of the latest and greatest and are willing to try just about anything if it portends significant advances in their niche domain, in their scope of work.
Jordan Schneider: Yeah, it's an interesting dynamic that we've seen with these labs, right? On the one hand, I can come up with crazy fun little applications, and there are power users who are building billion-dollar businesses with a headcount of three. But you also have these partnerships that all these labs are signing with consulting firms, with PE firms — because, no, you can't just press a button and have the AI solve everything for you today. There clearly needs to be this human layer of actual people to see the problems and solve them. I just found that to be interesting — that we're not quite here yet where you can, all right, GPT 7.0 or Claude Mythos 4.0 can just solve everything in government.
Barnett Koven: There are a couple of dynamics at play. First, these models are trained on what's available on the open internet, which is in many ways distinct from the niche missions of corporate customers as well as government, all of whom have very sensitive data that is not usually available in the public domain and available for training of these tools. So to think these tools out of the box can do everyday tasks that you and I do, like plan a grocery list — absolutely. That's not a perfect parallel to engaging in complex foreign policy negotiations per se.
The second piece is the garbage-in-garbage-out problem that's always been the case with data science, which is exacerbated here. Someone needs to have not only the institutional knowledge, but also know where the data lives, what limitations are evident in that extant data, and so on, in order to get the best value out of those models. And the final thing I'll say is metadata management has become huge here. Having clean, curated metadata gives an LLM some additional context to grab onto that massively improves the quality of the output.
Mirror Imaging in the Age of LLMs
Jordan Schneider: You mentioned AI for high-stakes foreign policy negotiations. One of the things I've been thinking about recently is what an eval would look like for that. Tactical-level decisions, you can kind of understand — you used to do COIN stuff. All right, if you give a model a whole lot of cameras, over time they can figure out who's leaving the house at weird hours with a gun on their shoulder. But if you're talking about foreign policy decision-making — like, what do you want to ensure is in the Ukraine peace deal?
The scary thing is these models are so good, and they seem like they can make your life so much easier, and they can write your Washington Post op-ed for you, that the temptation to help think through, okay, Putin gave me these 10 points, what are the 10 points I want to throw back, is going to be really high. But I wonder if you have any thoughts on what is the strategic-level negotiation eval you would want to throw at these models to actually have confidence that, no, this is better than my Russia desk officer or whatever — or that it can really upskill them.
Barnett Koven: The easy answer is to compare it against the humans. Give the Russia desk officer and the model the exact same prompts and see what they come back with, and compare across the results. If the model is coming back with similar conclusions to the Russia desk officer, one of two things is happening: they're both performing equally well, or they're both biased by the same things and performing equally poorly. This phenomenon of mirror imaging — right, we pay people in the US government all the time to go really deep on Russia, really deep on China, whoever the adversary du jour is, and yet they always do so from their own perspective as a US military officer and assume, at least subconsciously, that the system works the same in that quote-unquote competitor or adversary nation. I think that's problematic.
The alternative, though, is that when they diverge, then you can start interrogating why they diverge. If you're smart and you have the background information, you can trip up the model and see where it's making mistakes. Or you can say, aha, I actually learned something here, and the reasoning that was involved is really novel, and we didn't think of that, and we need to incorporate that.
Jordan Schneider: Yeah, it's interesting. I was actually having a conversation this morning with someone who was in the “deep state” for a long time and now is on the outside and gets to hire an international team — so has a Japanese national living in Tokyo as their Japan analyst. They're just like, look, it was a lot of fun, but I worked with Americans exclusively, and my entire inbox was full of Americans and their analysis of the world. So this idea of these models being able to inhabit different hats or different perspectives, which are not, like, you know, US citizens living in 2026, is an interesting one. If you're only using American models, is there something you're missing? Or if we're trained on the by-and-large English-language internet — that'll be a fun thing to tune, actually. Just like, okay, pretend you're a China desk officer, or an MSS analyst, or an Australian analyst. How do you think this stuff through?
All right, one more COIN thing before we jump into the BIS stuff. Is AI going to — can AI help you win an insurgency besides just doing targeting? And we weren't bad at targeting, right? That wasn't the problem.
Barnett Koven: I think the issue there was being able to separate the insurgents from the general population. If you could identify who the insurgents were, figuring out a solution, whether that was kinetic or otherwise, was not particularly challenging. I think it's a really interesting question. And I think one of the reasons why the US did reasonably well at counterinsurgency was the willingness of the US military to devolve decision-making authority down to junior officers and non-commissioned officers. That works because of the high educational background and level of training in the force, and this generalized degree of trust.
That being said, having machine learning models available on the edge, on tactical devices, I think could help with additional context and perhaps also be a sounding board for those JOs and NCOs as they're being asked to make incredibly difficult decisions rapidly with very limited context.
Jordan Schneider: Yeah, I mean, the thing about the Russia negotiation with Putin to end the Ukraine war is that's a one-of-one, right? You can't run an experiment about it. But okay, I have a thousand villages and 10,000 different officers running around, and someone's going to build the water treatment plant first, someone's going to invest in the police station — and you can learn from that over time in a given situation or given time period in a way in which I think the actual heads-of-state negotiation stuff is just going to be incredibly difficult. But okay, I'm a believer. Not to say we should start a war to test it out, but…
Unwrapping the Shell Companies
Okay, so let's come to some of the work that you guys have been doing with the federal government since you headed over to Accenture. Start with commerce, maybe. What is the dream of how data and AI can unlock better export controls enforcement?
Barnett Koven: Yeah. At least as it relates to your podcast and China, this topic du jour is export control policy. There are two sides to this coin. One: what should actually be controlled? And two: how effective are those controls? Noting that whether it's China or another state, you have a thinking opponent who is going to constantly improve on the ways in which they try to skirt those regulations and obtain whatever technology it is they're seeking. In this current moment, we're largely talking advanced semiconductors and to a lesser extent critical minerals — although I think China has done quite a good job at cornering the market on much of that latter supply chain.
Here, there are a couple of things we can do. One, some predictive modeling based on past behavior to try and identify higher-risk entities. The Office of Foreign Assets Control in the Treasury has done this for quite some time with respect to drug dealers and others, trying to identify what accounts to sanction next. The same logic could apply toward that targeting.
But the other aspect is unwrapping this maze of shell companies. First off, licit companies — especially Fortune 500 companies — have multiple subsidiaries for various legitimate legal and tax purposes. So shell companies in and of themselves are both very common but not necessarily indicative of nefarious behavior. One thing that we found really helpful is using entity analytics, like entity resolution, in order to try and make informed decisions about how to merge different publicly available as well as proprietary US government datasets, plus Bloomberg and S&P corporate records data, in order to create a master repository of the companies and their relationships to each other — to make it easier to identify that yes, in fact, this shell company that we've never heard of before is, through seven other corporate entities, connected to a procurement arm of the PLA.
Jordan Schneider: And is that — there are edge cases now where a frontier model can do the work that someone would have had to stare at the two things and be like, oh, this is the same company. Is that how this is getting accelerated these days?
Barnett Koven: That's correct. And you don't necessarily need gen AI for this — more traditional AI/ML can also be used. There are various nuanced questions you would want to ask in order to determine whether an LLM or a frontier model would actually make sense here. But in general, yes.
Jordan Schneider: Yeah. I guess the question is, what is the ratio of stuff where — you mentioned you guys have a Databricks partnership — what percentage of this stuff is just level one, this should have been done in 2016, but the new technology we have today is just making it cheaper and faster, versus, oh man, here's all this, we're catching people we never would have caught if it wasn't for the technology that's available today?
Barnett Koven: Historically, pre-the-last-four-or-five-years gen AI revolution, I would have told you that 80% of the work that a high-end data science team is actually doing is relatively run-of-the-mill data engineering — the cleaning and transforming of one's underlying data in order to enable the advanced analytics. I think a ton of that is still happening. The reality is, a lot of our data in government, but also in commercial, just isn't ready for AI/ML, let alone gen AI. That being said, maybe it's gone from 80% to 65% in the intervening five years. The LLMs themselves can be used for a lot of this rote ETL work as an accelerator, which also means there's more of an impetus for our customers, be they government or commercial, to actually do that annoying, painstaking data-cleansing work — because now I can do it at a third or a fourth the cost of what it was five years ago.
Jordan Schneider: So the ambition — you can just dream more with your budget at BIS, like I can accomplish a project that's three times as big as I could have three years ago.
Barnett Koven: 100%.
10x Engineers and Token Budgets
Jordan Schneider: I want to talk about talent. You guys are a federal contractor. We’ve had a lot of discussion on past ChinaTalk episodes about declining state capacity and the challenges — the need for government to be able to cook the food that it eats. Why should AFS exist? What’s the justification for the government relying on outside players to help it upskill in this regard?
Barnett Koven: The answer I'm going to give you is specific to AFS and just a handful of our competitors that have really robust commercial businesses — because I think that truly is the differentiator. Specifically, we have about 15,000 employees dedicated solely to serving the US federal government. We have 800,000 employees globally, principally serving the global Fortune 3000. We've all heard these biases that corporate America is ahead — or the corporate world in general is ahead of the US government in terms of innovation. That's a whole longer discussion that we could unpack. But the reality is the drivers and the incentive structures are quite different.
That being said, if you have 800,000 people serving the global Fortune 3000 who are innovating at a rapid clip, the ability to lift and shift those innovations into government, I think, is immensely valuable. Let the US government take the insights that we've developed over the years with our commercial clients and benefit from them from day one, as opposed to having to invest and solve these on their own. It also just gives us immense reach-back. If the government wants to do something new and novel and they can't hire a person as a Fed to do this because they only need the skill for two weeks, we have that novel person somewhere in our organization that can be brought in, and that can be productively used serving commercial customers most of the time. That flexibility is also part of the value.
Jordan Schneider: From a personnel perspective, it's interesting thinking about this, especially on the data engineering, data analyst, broadly software work side. You could have 10x engineers back in the before times, and now you can have like 100x or 1,000x engineers if there's someone who really understands how to burn tokens in a productive way
As you're interacting with different corners of the government, have you seen these people? Do they exist out there? What are the craziest things you see folks dreaming up, and how has that interacted with your work?
Barnett Koven: So we support DARPA and IARPA in other parts of the AFS ecosystem, and obviously those are the parts of the government that are doing this really wild experimentation. I've heard similar things with actually much higher numbers even, in terms of expected token consumption. That being said, I think where we ought to be heading for most of the enterprise use cases in government as well as established corporations is how do we get smarter about where we use generative AI and agentic AI. That doesn't mean just throwing tokens blindly at the problem, because that doesn't lead to a sustainable enterprise-scale solution. Rather, how can we bring in the traditional non-generative AI/ML that has been tried and true for decades — have that be 80 or 90% of the solution because it works and works well, and then you sprinkle on gen and agentic AI on top for niche aspects of the problem where traditional AI/ML is not going to be able to perform as well.
In doing so, you constrain token consumption, so you keep the total cost of ownership down, but you actually get better results by virtue of using deterministic processes as opposed to probabilistic ones for pieces of the problem that can be answered deterministically. And I think that's the special sauce — that's what a firm like AFS can bring to the table. That doesn't mean I'm not all for throwing everything at an LLM when we're in the really early experimental phases, way before we consider moving something to production. And that frankly does allow our teams as well as others to just be much more innovative and try things and fail — but do so faster and then learn from that failure.
Jordan Schneider: Yeah, you see CEOs now having to take into account, like, I'm not just hiring an engineer, I'm hiring an engineer plus whatever — 10 grand, 100 grand, 500K of capex that just goes along with this human being. And the answer is we don't know, right? And it's going to change over time. Is the federal budget process going to — you get headcounts oftentimes from Congress, right? Not necessarily compute budgets. Just speaking of BIS, there was this multi-year-long effort to try to get them $100 million more dollars. And, you know, if we're focused on executing the mission — is the mission 50 more case officers? Is the mission five case officers and a ton of models just spun up tracking every news story and every ship on the planet to figure out what's getting smuggled somewhere? I don't know. It's this sort of like, what is the optimal way to spend government money to get the outcome that you're looking for is probably the most wide-open that it's been in a really long time.
Barnett Koven: Absolutely. And the government budgeting process — realistically, your budget this year was decided two years ago, give or take. That is problematic here. It's not just the headcount that you were mentioning, but also what's your token consumption, what's your underlying compute bill from your cloud service provider. That is going to become harder and harder to predict given the rapid pace of innovation, the rapid pace of change.
I think the ways to solve this while still providing accountability are, one, provide a bit more fungibility between tokens for this model versus tokens for that model, tokens for this use case versus tokens for that use case, as long as they're tied to overarching priorities. And also, to aggregate across a department or agency — because it may very well be the case that this particular office in a given agency has a really novel use case that we couldn't have predicted, but you know what, this other department over there has a problem that's going to be much harder to simply solve away by throwing tokens at the problem. Having some fungibility within the overall department in terms of how that budget gets spent, with appropriate FinOps in place so that there is accountability — I am not suggesting we blindly spend taxpayer dollars or just remove constraints here — is going to be the direction the government has to go if they want to continue to leverage these latest and greatest innovations.
The Gun-Armor Race
Jordan Schneider: Do you want to talk a little bit about the national security piece of this? Because if you want to do all this stuff in a cleared data center or whatever, that's a whole other lead time, which isn't just shifting it from BIS to the CHIPS Act or whatever in commerce — you've got to build a place and buy the chips, which is not changing a line in a budget.
Barnett Koven: Yeah, absolutely. I had conversations in the last week with two different foundation model companies, both of whom were touting that, hey, we are releasing the latest and greatest models into our FedRAMP-but-unclassified US government environments on the same day they're becoming available to the public. The reality is, once you get to IL6+ — those classified environments — that is not the case, and that realistically can't be the case, given security constraints here.
That being said, it is very much an arms race. I had this amazing PhD student who was an Army cyber officer who came back for her PhD, and she looked at risk aversion in the AI era in government. Her most powerful insight was: inaction is in and of itself risky. We so often focus on buying down every risk for the action we're going to take, and we potentially don't take that action because of the risks we can see, that we forget about the fact that if we don't do something, there is not only a risk, but a guarantee that we will be left behind by peer and near-peer competitors who are less risk-averse than us and are willing to take risk after risk after risk. I do think we've seen that with China in particular.
Jordan Schneider: Can you talk a little bit about what it takes to bring something from an unclassified to a classified network? I mean, are you just worried that those models are going to run around and delete stuff, or is it just that the compute isn't there in the secret data center?
Barnett Koven: So the compute is actually less of a problem in classified environments. It's more what guardrails need to be in place. Is there any chance that any of this data could get out into the wild? And again, harder in a classified environment, but not impossible. Is there a risk that this model is going to inject something into our environment that could be deleterious? And frankly, has the model been poisoned in such a way that it's going to give us the wrong answers? There's been some amazing research on poisoning.
Unfortunately, this is a bit of a gun-armor race wherein the cost of the offense increases linearly over time, whereas the cost of the defense increases exponentially. There’s an inherent advantage to someone who's trying to do bad when we're talking classified spaces. So there's just a need to be especially cautious there.
The Manhattan Project Pose
Jordan Schneider: Let's talk a little about the Department of Energy. What is the Genesis Mission?
Barnett Koven: Yeah, absolutely. The Department of Energy announced the Genesis Mission, and there was this signing ceremony at the White House that our CEO, Ron Ash, participated in. Originally 26 different companies — so Accenture Federal, and then mostly foundation model companies, the hyperscalers, etc. — got together and signed this commitment to work with DOE and work with the US government in order to deliver on this promise.
The promise is, in effect, that DOE will become the leading organization across the US government for generative AI, and that we're going to treat advances in gen AI similar to — the way I've heard it described from some of our clients is — similar to the Manhattan Project, which is convenient since you have that poster back there for the 21st century. Provide the appropriate resourcing to make sure that the US remains preeminent in this domain, and importantly, ensure that we can safely build a moat around critical scientific discovery and other use cases to de-risk the application of gen AI. Because my point to you a moment ago was not that we shouldn't use these tools in a classified environment — just that more care is needed. In fact, not using them is the wrong answer, because again, to my risk comment.
Building the Backbone
Jordan Schneider: So how is AFS going to solve the problem?
Barnett Koven: What we're doing is we're building what we've colloquially termed the backbone for scientific research. This is the infrastructure that will allow secure collaboration across the labs, DOE headquarters, NNSA, critical industry and academic partners — with the tooling to pull in data from commercially available instrumentation that you would find in hard sciences research labs, equipment made by the likes of Siemens and what have you — and make it accessible with robust access controls, but accessible to those who have a need to know and for whom it would be valuable to their research enterprise.
The hard part now is not so much coming up with the latest and greatest foundational model — and frankly, there are foundation model companies that are doing amazing at that. It's how do you connect those models to a niche use case like scientific research and ensure they have access to the relevant training data and the relevant insights.
Jordan Schneider: Gotcha. Let's give an example. What is a thing that Sandia would be doing today that not enough people would be aware of, or an insight that would have disappeared without this in 2028? Pick your lab.
Barnett Koven: Pick your lab. One answer would be around supply chain security. One aspect of supply chain security is, where are there opportunities to change the way in which we produce critical minerals? Beryllium is a great example. It can actually be made from a number of different types of ore, some of which have not been commercially viable in the past, but maybe now. How much of that data was discarded 100 years ago because there was a cheaper, easier way to do it at the time that is now no longer possible because a strategic competitor has cornered the market for that mineral type? And then where is the data on what happens to exist buried in the ground in Utah and Nevada and what have you that might be relevant to solving this problem? And separately to those geological studies, where is the underlying data on novel approaches to adjusting how we treat tailings from existing mining operations in order to try and recover things that in the past were just waste minerals — because again, it didn't make sense or we didn't have the technology to actually extract them?
Jordan Schneider: So one of the reasons this exhaust fumes of scientific research hasn't gotten out is not just that you can only publish results with statistical significance, but also it's annoying and it's more work for the researchers themselves. Like, okay, if I only have to write up 12 things, now I have to upload 11 other datasets. What's the solution on that front?
Barnett Koven: Well, AI can write the reports for you. I mean, being a little bit glib here — and I do think there still needs to be a human in the loop here — but if it can do 80% of the work, why not? We're seeing this within data engineering. One thing that all of the data engineers, almost to a fault, that I supervise hate doing is documenting their code. But it's critically important so that the next person can pick off where they left off, and so we can interrogate, did they actually do what they think they did, or is there a mistake here that's potentially consequential? We've seen amazing strides from Claude Code and Codex in terms of doing that commenting-out or documentation of code for us.
The same logic applies really nicely to report writing. And frankly, there's an amazing corpus of text in terms of published peer-reviewed articles for what that should look like, such that it wouldn't be unreasonable to expect that a model could get you 80% of the way there.
Jordan Schneider: But if I'm doing physical experiments, it's important that I do all the steps, right, and remember all the steps. I mean, is the actual end state that I just have a camera in my lab that's taking notes and writing up everything I do over time? So if I only pour half a beaker in instead of a beaker, it'll notice and write that up and upload it to your thing, as well as let me know that there's some user error in here, which is why I didn't find the new beryllium compound?
Barnett Koven: It could be that. In reality, a lot of this research is a little bit more automated and whatnot, so it's not me hand-measuring, but rather using instrumentation to do that — to which the instrumentation is producing output that by and large is not being captured, or at least not retained. So I don't even think we have to install the physical cameras, but we just need to collect that data and actually do something with it.
Jordan Schneider: Cool. I'm excited for that. I think it's a net positive for humanity. We're going to find more cool stuff.
Can you talk about the secured versus not-secured part of this work?
Barnett Koven: Yeah, absolutely. One of the key things to make this palatable for research when it comes to critical minerals, nuclear energy, grid reliability — all aspects of, frankly, our competition with peer and near-peer competitors that are going to be consequential — is making sure that while this data is easily shareable amongst those we want to have access, there are robust guardrails in place.
These should be models that exist on existing secure cloud environments, or even perhaps bare metal. Our national labs have made major investments in high-performance compute, so they have the bare metal infrastructure that most corporations and others do not. But also implementing really robust RBAC and ABAC controls such that you can be extremely granular down to the row or column level — who should have access under what circumstances for what research. It would likely be the case that you'd want to share more widely within the national labs than you would with certain university partners, than you would with certain international university partners, and so on. That granularity is going to be critical to making this enterprise work.
Jordan Schneider: RBAC and ABAC — roles-based and attribute-based access controls. You want to explain what that is?
Barnett Koven: Yeah, absolutely. The idea here in lay terms is to make it really, really easy to decide that you, based on your specific role within an organization, should have access to a piece of data — or definitely should not have access. On government systems as well as most corporate systems, you are signing in with some type of either physical token or unique login and password and whatnot. So using that single sign-on, it will identify that you are this person, this is your role within the larger research enterprise, these are the projects you're working on, and so on — and automatically govern access. Importantly, according to the rules set in place by the individuals or organizations that generated that data.
And I think one critical piece to this beyond the technological is how do we get the buy-in from the various stakeholders. The critical piece there is going to be ensuring that ultimately they have discretion over how the data they generate is shared and can impart that knowledge. But then these RBAC and ABAC controls mean that you don't have to go email John over there and say, here's what I'm working on, can I please have access — and then John doesn't go from being a full-time research scientist to a full-time email-answerer and permissions-granter.
Jordan Schneider: But if I'm the Secretary of Energy and I'm cutting all the checks anyways, can I just say, sorry, you have to share this with everyone I think is responsible? What's the actual friction point?
Barnett Koven: Yeah, so in that case, the answer is nuanced — and I will caveat, I am not a lawyer. I am the kid of two lawyers, but that only gets me so far. Depending on when you sign a research grant or contract with the federal government, there are terms and conditions that you agree to, and depending on those terms, that will govern how the information is being used. Typically, though not exclusively, the government will retain rights to the research they funded for their own purposes. But that doesn't necessarily mean they can go ahead and share that with a third-party commercial entity or university partner or what have you. And again, it's going to be very dependent on the terms and conditions of an individual grant or contract.
Jordan Schneider: Gotcha. Are the labs excited about this? I don't know — it must be fun if I'm a scientist. Like, okay, if I'm only capturing 10% of my work now, just like now you're a 10x scientist and you don't even have to do anything else except hit upload every once in a while.
Barnett Koven: I think so. At this point, while we have viable demos and whatnot, we're not at the point where we're fielding this and they can get hands-on. So I think the excitement will really come the second people can get in and play, especially for scientists or just people like me who are naturally inquisitive. The ability to actually touch it and play with it is the real differentiator. I'm pretty excited about the concept, but I'm doubly excited because I've actually played with what we've been prototyping at AFS.
Koven's point that inaction carries its own risk lands hard on export control enforcement. BIS has always been a small office trying to police a huge volume of trade, and the PLA's habit of hiding behind layers of front companies turns every license review into a guessing game. Entity resolution that stitches public filings to government datasets could finally let investigators see the same owner behind ten different names. The poisoning worry seems just as important. If defense costs climb exponentially while offense climbs linearly, classified environments may end up leaning on smaller, tightly audited models. Curious whether Genesis builds in that kind of audit trail from day one.