Article · July 22, 2026

Behind the FDA Cybersecurity Guidance: What the FDA Guidance Actually Says

Beau Woods and Seth Carmody on the history of the FDA premarket and postmarket cybersecurity guidance - what went into it, how it is commonly misread, and how it was intended to be read.

A candid conversation and behind-the-scenes look at what went into the FDA premarket and postmarket cybersecurity guidance, from two of the people who shaped it. Beau Woods and Seth Carmody cover what went into the documents, what they are actually trying to do, how they are commonly misread, and how they were intended to be read and followed.

Guest: Seth Carmody, VP Regulatory Strategy at Medcrypt. Former FDA cybersecurity lead and architect of the postmarket cybersecurity guidance. Connect with Seth on LinkedIn.

Transcript

Beau: Hello and welcome to the live stream for behind the FDA Cybersecurity Guidance, where we will be taking you through a little bit of the history of what went into making the current guidance documents, what preceded them, what we were thinking as we sat in the the office and were belting things out on on the whiteboards. so hopefully many of you will know us on the stream. some of you will not, so we’ll introduce ourselves and then kick it off. my name is Bo Woods. I’m founder and CEO of Stratego Security, and I’ve been working in healthcare cybersecurity since 20, 2004, roughly, with medical device makers, healthcare providers, physicians, clinicians, patients, cybersecurity researchers, and and others, including a stint at the FTA, where I spent one year as an entrepreneur in residence with. This guy, Seth Carmedy. Seth, over to you.

Seth: What’s up? I thought we were doing a video game screen, so I’m a little disappointed that it’s regulatory focused. No, it’s all right, I’ll I’ll take care of it. Seth Carmody, VP of sales, which is a very strange title to hold. but I guess careers are weird. No no weirder career than mine for sure. We can dive into it if you like. so running sales at Medcrypt. Medcrypt’s products and services company that I stumbled upon after I left FDA. Spent a decade at the FDA. policy wonk. Is that is that sufficient, Bo? Do you think it sufficiently covers my my expertise at the FDA? I was a very bad reviewer. Yes, there we go. Got it. Medical device medical device reviewer mediocre. So there we are.

Beau: There you go. Out of out of how many reviewers at the FDA?

Seth: Seven hundred. I was probably six six ninety nine, in terms of rank. Yeah.

Beau: Six ninety nine. Nice. Nice. So as long as you got that one percent. Either way, if you’re the best one percent or the bottom one percent. okay. Well that’s good. That’s good to know.

Seth: There was somebody worse than me. I had to be, yeah. Every time I walked into my boss’s office, she was like, man, I remember w I was at a conference one time and I actually apologized to one of the regulatory affairs people at whatever company it was. I’m like, you got me in my second 510K. That just I I think straight away you deserve an apology. She laughed.

Beau: Well, and then you you got you did so well at that, you got promoted up to to doing other things, right?

Seth: Yeah, yeah, it’s it’s a it’s a weird transition. So think in twenty sixteen, twenty fifteen, something like that, I moved over to policy shop out of review. I think I I get bored after four years, so I spent four years as a a reviewer, and then four years at policy. So I I it’s like literally four years and two months. So I I don’t know what that significance is, forty two. but apparently I can only spend four years and two months in any given role.

Beau: And what did you do before you were a reviewer set?

Seth: actually and I I remember Bo would we would go back to origin stories. we w this was the the people were like, wait, tell me more, hit the tell me more button or whatever. It’s like, yeah, I started off as a chemist. a literally legit PhD chemist. I wrote a thesis, it wasn’t very good. and everything. And I stood in front of a group of folks who are also like PhD chemists and they’re like, This thesis isn’t great, but it’s good enough. So they They they gave me the PhD title. Studied at UCSD, UC San Diego. I still have my San Diego phone number. I I think it just looks cool. when people say 858, right? They’re like, this guy’s military or something. And then went to Indiana University and that’s where I finished.

Beau: Got it. So so chemist to reviewer to writing cybersecurity documentation policy for

Seth: Yeah, weird.

Beau: the FDA. It’s a it’s quite a transition. Yeah, I my degree is in psychology. And so I also have just a weird wandering, stumbling pathway to to getting here. But you know what? Whatever works. I think

Seth: What do what do you think what what was the linkage do you think between your degree and what was the stepping stone to the next thing?

Beau: Yeah, for me it was I always knew I was gonna be doing something computery because I’ve always been a geek. but I just found the coursework uninteresting. and hanging out with, some of the the hardcore nerds and geeks, I was like, I I don’t feel like I fit in. And so I I found some place that I actually enjoyed studying. the the coursework and where the people were cool and relaxed and more fit my vibe. and so I ended up getting a degree in that so that I could have four years of just enjoying the thing that I was studying before I went out and had to make money in the real world. jokes on me. ended up in in computers and and doing lots of various different things with them and breaking them from time to time. always for for good reasons.

Seth: You spent a spent as a consultant, yes. So you’re you’ve you’ve been there. You’ve done that.

Beau: Yeah. I so I I spent a few years at a hospital helping to unhack medical devices and build our security infrastructure. and then went out and did five years as just a a consultant on the road. Actually a couple of years I was on the road for about fifty weeks out of the year, going to all kinds of various different organizations and seeing how they did things and I learned a ton. and got

Seth: Yeah to

Beau: great airline status and hotel status, which was nice. and then came into the back office and started working with the the business teams on how to run a consulting company. And after a couple of years of that, decided I wanted to go do it for myself. and very quickly found out that it’s a lot harder than working for somebody else. and in ways that aren’t the the fun, cool you get to do. technology things. It’s always like, how do you set up a a business and get an accountant and and do that? So needless to say, when I ran into Suzanne and and one of her counterparts in 2014 at an Atlantic Council event and we were talking about healthcare cybersecurity and medical device cybersecurity and the the draft the initial draft pre-market guidance had just been put out And so I I think we spent about an hour or two talking, and then kept in touch, ended up talking a lot more and and this was just at the time when I I don’t know where you were in your fractured career arc in twenty ten to twenty thirteen. but that was around the time when there was a lot more security research happening on medical devices. This is when Barnaby Jack was doing some of his research.

Seth: Yes.

Beau: when Jay Radcliffe was demonstrating things. And so the the kind of vibe that I got from Suzanne and team at the time was like there’s a there there. We just don’t know what it is yet. And we need to figure it out because it’s it’s critical for our role in protecting public health and the public health system and patients and for safety and effectiveness. the cybersecurity is a vehicle to safety and effectiveness. So how do we get to understand that better, get to understand the people who are doing it better. and so I remember holding a a small conversation. It was a handful of cybersecurity researchers. and Suzanne, I think Bakool was there, and I

Seth: Mm.

Beau: don’t remember who else was was in the room, maybe Afton. and we were it was a half day and we were walking through security research this, that. And I I think it was Bakool who just seemed I don’t know, somebody in the room, it seemed like they just couldn’t get their head wrapped around who security researchers are. And they heard security research and they thought hacker and they thought criminal. And so we we understood that we actually had to break it down and say, here’s why security researchers do what they do. Here’s why they’ll, go and get a device and set it on the countertop and Pull it apart to find the bugs and put it back together and not be able to sit around the table with their family for a month because they’re they’re too busy poking and investigating. and and that unlocked something. And once once [FDA] understood that security researchers were here to help, they just didn’t have the vernacular of speaking with medical device makers or the FDA, then all of a sudden it was about how do we translate. Between each other’s languages, so that we can serve public health better together. And so that we can harness the best of technology to serve this purpose and this cause, rather than inadvertently opening the door for accidents and adversaries to have some negative impact on the thing that we we want to try and achieve. coming out coming in as a total outsider to the regulated space, but being good enough at at having that kind of mission drive and vision foreseeing that we could work together and just collaborating and and building something together with

Seth: Obviously it had an impact ‘cause at some point I remember there was a headline, Jeff Surin says, hire a hacker, right? So it was a good I don’t know if he got the message quite right, but it was a a good one to to grab attention for sure. but we we all understood what he was saying, right? Yeah, go go hire somebody that has an adversarial mindset. And I I think it was important at that point, especially in the early policy journey, because and I as you were talking, I’m starting to remember stuff. I’ve forgotten so much over the years, which is funny. But I remember thinking healthcare are peop are builders. They’re usually like constructivists, not to say the hackers are not. So they think about how can I put features together and stream together to have some, intended, health impact. They’re builders, right? And they don’t think in terms of b breaking something. They don’t think of how this thing can necessarily be used in a way that it’s not intended. There was a a quote from Dr. Eugene Vasserman, I’ll call him the prof. He hates that. but he said, it’s not what it’s designed to do, it’s what it can be mi made to do. which I thought was a a really great way to to s to, change the thought process when you’re working with manufacturers to that okay, I got you. If I’m not preventing it from behaving this way, then it it could be made to behave this way.

Beau: Right. Yeah, yeah. I’ll I’ll never forget. I think I’m pretty sure it was Christian DeMef who who said, as we seek to treat existing pathologies, we should take great care not to inadvertently create new pathologies. In other words, if we’re if we’re importing computing technology into all of the stuff that we’re using to save lives, that computing technology has pathologies all its own. It has failure modes all its own. It has Hazard conditions that are unique to it. And if we’re not really careful, we could inadvertently import those in ways that go untreated. And so I think it’s, finding those ways to speak with, speak across communities in ways that they can both understand is extremely hard to do. But once you can do it and get them aligned, it really helps unlock some fantastic things, like some of the best conversations and and breakthroughs that I’ve had talking with medical device makers or regulators is when you can get them to drop into that adversarial mindset and then use the knowledge that they already have to figure out the implications of it. Right. in ways that I certainly can’t, because I don’t have that deep base in biomedical engineering, for instance. But if I get a biomedical engineer to say, what could I do if I were trying to cause harm with this? And what are the safeguards in place for that? Then all of a sudden they just run wild and they find all the doors that they need to go close to the adversaries or the accidents. and and that can really open up and unlock something. And I think that’s one of the reasons why, Bringing it on topic. That’s one of the reasons why something like threat modeling is so critical and why the FDA has been pretty insistent in declaring that you need to do threat modeling. I remember sitting in a Northern Virginia conference room for two days with with you and a bunch of other people from the industry, working on a whole threat modeling workshop and and having some threat modeling experts come in and just walk through the process. and translate it. And that eventually became an MDIC document, which I know the FDA refers to a lot. It’s like the threat threat modeling for medical devices guidebook or handbook.

Seth: Yeah, playbook. Yeah, it’s interesting how things work out. It’s like I remember Suzanne Schwartz came in to my and remember my office? You remember I had my fake office?

Beau: Your office, yes.

Seth: Yeah, it was like the decoy office and it had my name on it and it it gave the air that I was would be there at some point after I came back from meeting, but I never went there. I went to the secret policy bunker with tons of whiteboards and materials to create policy. Yeah. I remember one day, well Suzanne knew I was there, right? Of course your boss should know probably where where you are. And then

Beau: Right.

Seth: she poked her head and she’s What do you think MDIC should work on? I was like, threat modeling? This I I can’t remember the year, but it was like it was it was really early. it this is early on. It might have been twenty seventeen, maybe. we’re talking I I think people just

Beau: Probably.

Seth: started to care about cyber medical base cybersecurity, medical base manufacturers as enforcement’s gone up in twenty twenty five ish. So that’s T zero in my mind. So we’re talking Ancient history. So twenty seventeen ish that conversation happened. And that’s why that that full workshop and MDIC effort and playbook came out.

Beau: Yeah, and remember because the the MDIC workshop was twenty twenty or twenty twenty one or twenty twenty two. It’s it’s been a while. Yeah. Yeah.

Seth: I think it was right at the yeah. it yeah, and then the origin story I was thinking back to this, I remember like you mentioned you you met Suzanne and and Co. around twenty fourteen, and the draft pre market had come out. I remember my role in that policy document was I showed up to the meeting, I tried to figure out how the hell I made it into that meeting. why was they that I’m like I’m a chemist and I I’m interested in software. something some terrible decision has been made like how am I here? And just remember listening to the conversation. I’m like, this is this is just really interesting. Brian Fitzgerald, I think, if if folks remember Brian was the had a lot of thoughts on cybersecurity, the proto thoughts on policy, and he was the one that was, really driving a lot of the the the aspects there, melding that with some of the s the sort of researcher community and things that we were getting from the researching community. that that ultimately resulted in that that first document. But I can’t say that I I wrote a lick of it. so I was just there hanging out and just vibing. So that was a long

Beau: Yeah.

Seth: time ago as well.

Beau: was. But then you did end up picking up the pen for that, right? How how did that happen?

Seth: I don’t remember what they I think Nick Thacker, I don’t know if folks remember Nick Thacker.

Beau: Yeah, yeah, yeah.

Seth: he spent a l of time in Suzanne’s group. By the way, Suzanne’s group is an emergency preparedness and response group. they responded to things like Ebola. And I think it which would have sudden seemed odd maybe in the beginning, but I think it was a really astute observation that cybersecurity was very much like that at the at the moment, especially in that reactive phase. but I I don’t honestly don’t remember. I think Nack Nick Thacker had probably taken all the comments from the draft and and helped finalize that 2014 pre market document. But I did I did steal the pen away from John Murray for the post market guidance. he went on vacation, for two months, because he’s just like, I’m I’m going on a cruise in Europe and I’m I’ll see you in two months. And Suzanne was looking at looking at the whole cybersecurity working group and she’s Well, this can’t stopped development for two months, I was like, Yeah sure, whatever, I’ll I’ll do it. So John John’s a a great guy and obviously I didn’t steal it from him. I just like again and he got back and he’s like, Yeah that makes sense. Let’s just do that.

Beau: Yeah. Yeah. I liked working with John. he had some really good stories about what came before and how things happened. And I remember in two thousand four, two thousand five, something like that, we had I was working on a medical device that had malicious software on it and trying to Get the manufacturer to fix the the software vulnerability. It was some it ran on Windows, right? It was some missing Windows patch that allowed malicious software to get in. And then it just caused a reboot cycle on the device. so it was effectively unusable. and I remember running across a document that I found out later John Murray had written, which was the original off-the-shelf software guidance that said something like, if you if you have software. not only are you allowed to update, we expect you to update it and to allow common protections that the hospitals need to put into a malicious software protection. And when I pointed this out to the medical device makers, they’re like blown away that this existed, right? We’ve come a long way since then. But I remember that was the the first exposure that I really had to the idea that regulated medical devices, somebody had actually thought about software. And that’s 20 plus years ago now. So it’s, like you said, a lot of people started paying attention around 20, 2024, 2025. But it’s been a long, long journey to even get there. And I think, one of the reasons why we we just dove into this conversation, which is great. But one of the reasons why we wanted to have it is because so many people are not aware or they’re forgetting about some of the history and the legacy and the things that have gone into. getting to the point where we are today where we can have discussions about threat modeling and medical devices, where we can have discussions about, what is an appropriate amount of safeguards on the authentication mechanisms for implantable devices when you want to connect and try and update their programming and things like that. so I think it’s a a testament to the work that the FDA and many, many medical device makers as well, a lot of a lot of really great people early on were trying to promote this idea that cybersecurity is patient safety within medical device makers. And I think a lot of it’s come together and stuck. And there is a a wandering path to how we got here, like you and I did

Seth: Mm.

Beau: in our careers. but a lot of those steps are important. And I think if you don’t know what went into the documentation and the the requirements and the guidance, it’s hard to know how to really meet it. and while the FDA has done a good job of going out on the speaking circuit whenever they release a new document and talking about it and what went into it, a lot of the decisions and some of those records and and some of the rationale isn’t transparent and readily available to everybody who is making a medical device, right?

Seth: We we lived every word, Bo. We lived every word and every word has a story, right? and we we begged borrowed and stealed stole content wherever we could, consulted with experts, but really, my job as a policy wonk was really a a synthesizer, which is weird because that’s what I did in graduate school. As a synth synthetic chemist, I would build molecules, so I just turned that attention and focus too. policy because I I just didn’t have the technology. I didn’t have the I didn’t have the the I didn’t have the background in security. So I had to go out and and get people’s opinions who who had that background and then stitch it together into a cohesive framework. And it’s it’s interesting you mentioned it’s wandering and arrived at it. I like the idea I love the call back to security and safety because I remember for a while, and I even made a slide like this probably in twenty fourteen where I had the the MasterCard logo. Venn diagram. safety and security is set two separate things and they overlap and in the middle. Sometimes those are there are things that are overlapping in the middle. And then I remember this epiphany. I was looking at it’s late in my policy days at FDA and I was with Suzanne. I can’t remember exactly. I think we’re writing the proto twenty two five twenty-four B legislation. And we look at each other and we’re like, we’re recalling all of our learnings. And we look at each other and we’re like, Wait, no, security is safety. Right? We just

Beau: Mm.

Seth: and and I think we had that pi hypothesis. But it took us twenty eleven to twenty nineteen to have the proof and data to be confident, write it down and back it up and be able to, argue for it. which which was a it was an it was a profound moment and and one that I think shaped once we knew that that t to be true, even though I had left the FDA shortly after that, that truth was was true inside FDA and they and they took that and ran with it. So ri really important data point to to have supported with evidence.

Beau: Yeah. Yeah. I’m I’m watching the chat over here. Somebody just dropped in they can remember the mantra that got repeated, which is look, we’re we’re an FDA regulated medical device. We can’t patch or modify it without FDA approval. and it’s because that the mentality for a long, long time was safety and effectiveness are antithetical to cybersecurity. Without understanding cybersecurity people didn’t understand the safety world, safety people didn’t understand the cybersecurity world, except for a handful of people back then. and I think the the interesting thing when I ask myself why is the why is that the case is like most of my career in cybersecurity I spent doing protecting confidentiality of data against threats. So

Seth: Mm-hmm.

Beau: most of the toolbooks are built around that. Almost everything, if you go out and you get the ISO framework for whatever, or the NIST framework for doing this, they almost all prioritize confidentiality of information above everything else. some availability is IT’s concern, but confidentiality of data. But in safety critical systems, the the bit that has to flip is you’re looking at protecting the integrity and availability. To close that CIA triad in cybersecurity, integrity and availability of human life and public safety. And that is a different playbook. That’s a different set of tools. That’s a different set of consequences. Because if you look at the way that most medical device makers see cybersecurity’s being done, it’s something like, well, you got your password wrong three times, therefore we’re going to lock you out for an hour. And if you’re in an emergency department, that’s catastrophic, right?

Seth: Yeah.

Beau: That just can’t happen. That has real world consequences that are not about privacy. They’re not on in intellectual property or credit card numbers or anything like that. It’s actually people coming to harm. It’s it’s people who have degraded lifespans or degraded quality of life because you have delayed care and treatment in, certainly some of the most urgent cases, heart attacks, strokes, and other things where that timeline really matters. So I think that looking back, I think medical device makers weren’t wrong. The way that they had seen security dude been done in the past and been told to them in the past, was at odds with safety and effectiveness. It just took some concerted effort from people who believed that they were not at odds with each other to work out some better ways to do it and to learn from other areas too, right? Medical devices are one example of safety critical equipment, but you have airplanes, which have been doing this for a long time. You’ve got cars, you’ve got some of the other things where they became connected on different timelines and different time scales. and so drawing from those, building something that’s fit for purpose for medical devices, and then going out and continually refining and improving it, I think that’s what the FDA has done an amazing job of doing over the last 20 plus years.

Seth: it’s interesting. I is listening to you talk. I was remembering these slides that I I think I baked soup shortly after I left. And it helped solidify, the the axioms that we uncovered during my time there. one of them was, and I’m trying to thread the needle here, I haven’t thought about this in a long time, so I’m gonna try to I’ll give you the I’m trying to do it in a minute and fifteen seconds, so time me. the the whole bar to market, right, was it’s reasonably safe and effective. It’s not 100% safety, it’s not a hundred percent efficacy, right? So there’s inherently risk i inside this contrivance, right? Which is I think a word they use in the law, which is a fantastic word. I don’t think I ever got to use it in my policy days. and what I was thinking was, well, you provide an evidentiary basis that establishes a static sort of risk profile, right? we we can say that we have a reasonably safe and efficacious because we have done our due diligence and we have characterized this surface, this risk surface. And what security ultimately brought was it it brought an uncertainty, or brought an uncertainty to that di that dynamic relationship between benefit and risk and tipped the balance towards risk unless that we could demonstratively pro prove or show. that we could maintain safety and efficacy while under duress. and a a graphic would be amazing right here. So I’m not but I’m not gonna go go that far and draw it. but it it actually for me is that quintessential point about security’s relationship to safety and efficacy and how it really upended the the model that FDA’s operating. but shows how flexible and creative they were in, adjusting to that that equation.

Beau: Yeah. I I I think that’s a great point. And one of the things that I like about reading through even some of the early software documents, I’ve traced this back to nineteen ninety-nine, some of the documentation from there. And they’ve continually made a point that likelihood should not be an indicator of whether or not you do something for software. And one of the ways that they phrased it, and I don’t remember what document this is, but it was something like. we assert that if an error has occurred, it will recur. And therefore,

Seth: Yeah.

Beau: you should treat it as if it will be a recurring thing and address it. And I think that that’s really, really interesting as a concept. because we don’t typically in software, we don’t typically talk about that as much. But when you have these types of interactions that can carry some risk of. whatever failure mode manifesting that that you’re thinking about. and you repeat that interaction dozens, hundreds, thousands, millions of times a year, of course, even something that has a very, very tiny risk or likelihood of recurring just by accidents will happen a few a handful of times, right? And so leaning into that, One of the ways I describe adversarial action is you take something that’s very, very unlikely to happen and you make it reliably repeatable at your own discretion, right? whenever the adversary wants, they can trigger this really weird rear condition that almost never would happen. And you might see a one every billion

Seth: Mm-hmm.

Beau: interactions, and you make it one out of every three interactions. And it’s like, okay, well, that changes the the likelihood equation completely. But That’s all at the discretion of the adversary. That’s all at their determination, at their whim. And you can’t really rely on the the kindness of known bad actors when you’re talking about public health and public safety.

Seth: Yeah. right on. I I wanted to I are you the host? Who’s the host here? Can I can I act as the host for a second?

Beau: I think Go for it. You’re the host. You’re the host now, dog.

Seth: I I wanted to interject a word into the conversation and we’re we’re gonna discuss. So it’s like remember Linda Richmond from SNL? I’ll give you a topic.

Beau: okay, yeah. Discuss.

Seth: Rhode Island is neither a road nor an island. Discuss. Cyber Secu

Beau: Coffee talk.

Seth: Cybersecurity Bill of Materials. Do you remember that policy gem?

Beau: I do, I do. That was a fun one.

Seth: I remember having this interesting conversation with Michelle Jump, I think now the CEO at MedSec. And she’s like, Well, Seth, what’s a list of software, what’s a cy what’s a software build materials? And it’s a list of software. And she’s So a cybersecurity build materials is a list of cybersecurity. I was like, that’s a fair point. let me think about that for a second, which there’s a whole we could probably do a podcast slash this streaming thing for software build materials to unto itself. But what I’ll say is the idea of sound. So when you’re thinking about policy, you’re like, okay, well, does hardware count? Right? You’re so you’re pre- you just sit back and you’re like, remember Brian Fitzgerald would sit back like this, and he’d be like, hmm. so you sit, you get you get very Brian Fitzgerald about it, and you’re like, yeah, does hardware count? You’re like, well, by the way, this was Meltdown Inspector was occurring at the time. So I blame that. It’s like, well, yeah, I guess. The the chip matters? the I think ultimately it was the software firmware that was flowing through the chip. Speculative execution stuff that that ultimately made that possible, but it’s like I guess the hardware matters. Okay, cool. Does the M6 screw matter? Okay, I guess that doesn’t matter. So what hardware has cybersecurity in it and what d like what so yeah, you can see the

Beau: But if you can take the six screw out and get access to the Ethernet port.

Seth: Are you a policy wong too?

Beau: Yeah, it’s a great question. I ha I have been called such. Yeah. and I think that that Spectre

Seth: Yeah.

Beau: and Meltdown is a really interesting example because you have a case where, yesterday the threat was only theoretical. you’re gonna exploit some speculative execution and the hardware. Well, yeah, I guess maybe that’s possible, but you’d have to do this and that and the other thing. You’d have to jump through so many hoops and then Spectre and meltdown was announced. And within a week or two, you could literally get a JavaScript framework that would run in your browser that would exploit this. So within the time span of less than a month, it went from theoretically, hypothetically possible, but probably would never happen, to yeah, that happens all the time. And so this is where I feel like the the type of risk equations that you tend to rely on in most of public health policy making break down. They have assumptions in them that break down because you can literally go from no knowledge or or trust that it will never happen to trust that it is certain to happen within such a short time span that you can’t adapt. And I think that this is why, we have such a a a hard time scoring risk because there’s such a temporal distinction to it. the risk that you have today is negligible. The risk that you might have tomorrow might be catastrophic. how do you even calculate that in a way that makes sense without some kind of a continual process, without re-ranking and rescoring things on a daily basis. And you just that’s an almost impossibility. even If you’ve got a bunch of tools and and tool sets and it’s like, all right, great, we’re going to use all of those things and software is going to do it or LLMs are going to do it. And that that’s great. But it’s still such a a static thing that it will not keep pace with adversarial changes or with new techniques or with the technology and the rate of change that happens there, new hardware, new software. so it it’s such a a hard problem, which is what makes it fun in a lot of ways is like thinking through and anticipating those types of things. But it makes it really, really daunting to do any kind of public policy around it when the lifespan is going to be more than a year or two. I think it’s why by and large, the standards world isn’t able to do the things in cybersecurity than it is for biocompatibility or electromagnetic interference testing or or those types of things. and so I think it’s a it is certainly a mindset shift. And it is something that I think the FDA has done a fantastic job. if you look not just across the healthcare world, but at all other regulated industries around the world, I think the FDA has done an exemplary job of being able to get ahead of those types of things, to build the ramp in a sane and sustainable manner of compliance and getting up to speed, of allowing it to be flexible enough to adapt as needed, but also muscular enough to force the industry to adapt these things so that they can stay ahead of of anticipated threats and reduce the risk aperture to a reasonable amount. in line with, like you said, the the reasonable, reasonably safe and effective concept.

Seth: That’s an interesting point. I I do think that if I were to go back and wave a magic wand over all all the policies that I worked on, I do think I would build in this notion of a maturity model, right? crawl, walk, run. I know you guys talk about that a lot. just to to emphasize where stuff’s hard, right? we we we deal with SBOM a lot in in in the day-to-day. And what a beautiful piece of policy. Hey, know what’s in your Know what software you’re using. Okay. n know what vulner the vulnerabilities against that software. Okay, we’re dude, we’re like batting a hundred here, a thousand. We’re we’re really good. and then, tell me what the what it like what what the risk is. Fantastic. and then what are you gonna do about it? Wow, that is just I do say so myself, but a very stunning piece of policy work. Really. Just Round of applause. come on now. In practice it is the hardest thing I’ve ever seen. we it it’s crazy hard for reasons that you were just outlining. It’s like we’re we’re not even there yet. we’re not even able to do it to that to that in that s in the spirit, even when people want to. It’s it’s really

Beau: Yeah.

Seth: interesting.

Beau: Yeah. We could so the S Bomb History podcast will be a whole different thing. and

Seth: Ha ha ha

Beau: and that one literally will go on for a week at a time, depending on who we we talk with it about it with.

Seth: Bo, did we settle on a format yet? Did what do we what do we settle on? Do we remember early days? Is it s gonna be Swids? Do we

Beau: Yeah, the the format is chaos. Whenever we feel like doing one.

Seth: Is it Swids? I don’t know. Who knows? Let’s figure it out.

Beau: yeah, no, the thing about S Bomb is in theory it’s easy. In practice, it’s a little bit harder. in practice with legacy systems, it’s almost impossible. There’s a range of possibility. I remember, you you and you and I were working on the software as a medical device pre certification program, which I think was a A brilliant and fantastic idea. How do you create public policy in a sprint agile mindset, taking advantage of the capabilities that exist when you are software alone and when you’re a probably always internet connected, right? It’s like great, we can see those as weaknesses or strengths, but let’s lean into the strengths and see what we can build. and I remember talking with some men manufacturers and they would say things like, you you can’t know what’s in your software. you can’t. generate a software bill of materials. It’s just, it’s impossible to do. and no one in the history of mankind has ever been able to do that. That’s you can’t do it. It’s an impossibility, right? And then talking to on the other hand, some other manufacturers, they’re like, how can you build software without knowing what goes into it? Wait, explain this concept to me. Why why do people think that that is not a thing? we we just accidentally as offcassing generate software bill of materials all the time. for quality reasons, for, reliability and normal software mechanisms. how do you have repeatable builds if you can’t if you can’t know it? And so this diametric opposition, but I think they’re both right in some old ways of thinking about it, you made the software once and then you send it out. And that’s a very nineteen seventies, nineteen eighties way of doing things. And then it’s like, well, you make the software and you make updates for it and you’re gonna do these things. So you need to keep around the engineers that knew how to do that because their development machine was the only one that had the the development environment that could actually reproduce this. And then you get it out the door and don’t tinker with those machines because they might just break. And, everything that they know was written down in a a static text document that we PDF’d up so we’d have it just in case, They won the lottery and weren’t going work here anymore. and then of course, through seven different acquisitions in 20 years, they retire and go somewhere else. And all you’ve got is the PDF. So, of course, you can’t generate a software bill of materials for those old things. But for anything new, again, you just like this is the way that it happens. I threw my code on GitHub and GitHub automatically built a dependency graph that you know gets me 60 to 80% of the way there. it’s like it just becomes easier with time and and with better technologies, better capabilities and learning those lessons. And so I think that SBOM is a great example of where the FDA saw the capability existed. It just wasn’t always practiced and said it’s such a game changer in terms of being able to identify systemic vulnerabilities that affect the entire supply chain or the entire ecosystem for a certain product. it’s such a game changer in our ability to understand complexity and dependencies and potential interactions at scale that we want to know what’s in there. And by forcing the industry to take that policy step, then you start generating the practices and the tools that make it just easy in what you do, right?

Seth: think it’s a great point. It’s is it physically impossible to do? The answer is no. And two, is it worth doing? Yes. Okay, so FDA’s job is to get people moving in that direction. And I remember, there’s plenty, there’s plenty of examples. for for the reason for reasons two that are coming to mind is that Rios, who I think Rios because he was he was there. he was there. He’s

Beau: Yeah. Yeah.

Seth: he he did some research on the Alaris punk. And the headline was 1400 vulnerabilities. I think sixteen. Somebody’s gotta fact check me on that. I think it was one thousand four hundred and sixteen. Let’s see if my memory serves me. This is what, twenty thirteen, fourteen, something like that. and he said, Well, well, what what’s the update? what what what’s the fix? It’s like two. Java updates. Okay, cool. So we should know what’s in it. And then one of the biggest hammers we like, we can take care of a lot of this attack service with two updates. It’s like, okay, that that seems like it should make it into policy someplace, right? And then the

Beau: Yeah. Yeah.

Seth: last one was Wanakry, which was a very interesting time to be a public health re regulator and policy wonk. and and that really showed me. how important it was to even take that further and and say, how do we connect this information to the people that are consuming the technology, right? The delivery side. And I will say, that’s still a frontier. from everything that I’ve been involved with, getting that information out and actionable in in the delivery side is still still in its infancy. Doesn’t mean that we shouldn’t be doing it, right? But we should still, keep in mind where we are in that journey, even though it’s been been a while. I thought this would be done in five years by the way. It’s going on twenty something, so I I don’t know how much time it’s gonna take to get

Beau: Yeah.

Seth: over the finish line.

Beau: Do you mean do you mean the the conversation around S Bomb or this podcast?

Seth: both? I don’t know. Yeah. No, just it w I didn’t know it at the time, but I think it and it’s been tremendously rewarding. It’s I’ve been a on a part of the a piece of societal change, right? This has been a part, this is a societal change level thing. And I I I don’t know about you, but I had never been a part of one. And I was like, well, this should be two take two years, right? And I remember looking at Kevin Fu and he and he’s like, What are you nuts? It’s like fifty years. Easy. So We’re Kevin Fu if you’re you’re on the betting platforms, you might want to lay down a bet for societal societal change. I thought he’s probably more closer to the the truth.

Beau: Yeah. Well, I I think we’ve gotten in all of our all of our name dropping by now. and somebody else did mention Kevin in the in the

Seth: sweet.

Beau: the show notes and some of his early research in tw two thousand seven or eight. Yeah.

Seth: Twenty eight. Yeah. Two thousand eight. Yeah.

Beau: On on implanted cardiac devices. But I think we are we are running over time. Our our produ our mysterious producer who may or may not exist is is giving us the symbolic cut symbol. So

Seth: Wrap it up.

Beau: I think we ought to wrap it up. so thank you, Seth, for being on. This has been amazing. Hopefully folks who joined in liked it too. We didn’t even talk about mythos. We didn’t even talk about a lot of the things we wanted to get to. so I don’t know, maybe we’ll we’ll have to do another one of these.

Seth: I I’m I’m game. Thanks for having me. There’s everybody.

Beau: All right. Cool. Thank you everybody for stopping by. And thank you, Seth, and we will see you again very, very soon.

Seth: Take care. Bye.

Get in touch

Working through the same questions on your device or program?