UL CEO Jennifer Scanlon on product security, counterfeiting, and AI

Immediately, I’m speaking with Jennifer Scanlon, who’s the CEO of UL Options. That’s Underwriters Laboratories – you recognize, the UL emblem listed on all of your electronics? That image means it’s been examined and located secure in a wide range of methods. UL’s been round for 100 years. It began as a means for insurance coverage corporations to do fireplace and security testing on electrical merchandise simply as electrical energy was coming into houses.

However now it’s in all places, and it’s a kind of corporations we actually prefer to poke at right here on Decoder that’s mainly hidden in plain sight — that emblem is on the whole lot. However scratch the floor and the enterprise of UL is fairly difficult. There are a ton of low-cost electronics on Amazon, and possibly folks simply care about value and never certifications. The corporate can be now making an attempt to do security testing for AI programs; it simply rolled out a brand new normal referred to as UL 3115, “a structured framework to guage AI-based merchandise earlier than and through deployment.” That sort of normal requires a whole lot of corporations and regulators to purchase in — and for there to be a solution to even reliably security check AI in any respect. After which there’s the construction of UL, which — properly, you’ll see. It’s difficult.

However positive, construction and no matter, we’ll get there, however first, I needed to ask Jennifer if she acquired to observe stuff explode within the testing labs. As a result of to me that looks like the most effective a part of working for a corporation that units security requirements. Lots of stuff blows up within the labs, and also you’ll hear Jennifer say her workplace typically rattles due to it.

However there are different problems: Proper on the tail finish of the Biden administration, UL acquired tapped to be the lead administrator for a brand new Cyber Security program that was purported to set an ordinary for linked gadgets — the entire Web of Issues. However then the Trump administration got here to energy, and good previous Brendan Carr has been arising with causes — which after all by no means really get articulated to anybody — why any firm associated to China is someway now a menace. That, apparently, consists of UL, which after all has security labs in China, since that’s the place the electronics are made. So UL misplaced out on that deal. I requested Jennifer about it fairly instantly, since that’s actually a microcosm of just about the whole lot occurring with security, tech, and China proper now.

There’s lots occurring on this one; I like after we get to carry hidden programs to mild. I believe you’re going to love it.

Okay: UL CEO Jennifer Scanlon. Right here we go.

This interview has been calmly edited for size and readability.

Jennifer Scanlon, you’re the president and CEO of UL Options. Welcome to Decoder.

Thanks, Nilay. It’s such a pleasure to be right here.

I’m excited to speak to you. A few of my favourite episodes are after we demystify a factor that everybody takes with no consideration and the UL emblem is a kind of issues.

Completely. The UL mark is on billions of merchandise, and but in all places I’m going, folks have a look at me and say, “What precisely does UL do?”

Properly, my understanding is that you just simply drop issues off of cliffs and see in the event that they explode. Is that your day-to-day?

We do have individuals who drop issues off of cliffs and see in the event that they explode, however actually each single day, we’ve 15,000 staff around the globe working for a safer world. They’re testing, inspecting, and certifying merchandise. They’re additionally creating software program to assist our prospects handle their threat and compliance environments.

You run an enormous testing facility. Describe among the exams which might be finished and who will get to do them and what among the wildest exams that you just do are.

I all the time prefer to say we break issues, we blow them up, we mild them on fireplace. If you happen to had been to stroll into our testing facility right here in Northbrook, Illinois, in Europe, in China, in India, anyplace on the planet, you’ll first see massive electrical panels which might be there charging and discharging merchandise, batteries, and seeing what fails. Watching a lithium-ion battery the scale of my thumb blow up is fairly terrifying. It’s wonderful how large that blast will go. So we do a whole lot of inherently unsafe issues to check product security.

My most favourite check, I wasn’t there, however I acquired to see photos of it. We stacked two million soda pop cans in our large-scale fireplace testing warehouse after which dropped a lighted piece of paper within the center to see what would occur. And to this present day, I don’t know if we had been testing the aluminum, the labels, the contents, however I do know the exams failed. They had been purported to collapse and sort of collapse upon themselves and so they as a substitute exploded, and it took a variety of days to scrub up the 2 million failed soda pop cans. That’s what we do. We defend our prospects. They wanted to know that what they thought was going to occur didn’t occur.

Oh no. What’s probably the most harmful check that you just’ve gotten to be there in individual for?

Our hazardous location testing is in Northbrook and my workplace is true above it. Each occasionally you’ll really feel a bit of shake. And you actually assume, how dangerous might or not it’s {that a} lumineer in a flamable mud surroundings sparks? Properly, if you consider that, you’re out on an oil rig, you’re out in some manufacturing unit, a whole lot of lives may very well be misplaced. So whereas the check itself is properly managed, it actually makes you consider the lives which might be at stake with what we’re doing each single day.

Do you ever bail out of boring conferences and simply go blow stuff up for enjoyable? I’d completely try this.

I don’t assume the engineers would let me. However they do take pleasure in it once I come go to as a result of I do ask a whole lot of questions and I’m all the time fascinated by the brand new issues that we do.

I believe you need to ask them. I believe they might allow you to. I’ve acquired to be trustworthy with you. I do know a couple of engineers. I believe they is perhaps like, “Yeah, we’ll set one thing up for you.”

Fifteen thousand staff, that’s lots. The corporate began a very long time in the past as Underwriters Laboratories. Hearth insurance coverage corporations wanted to verify electrical gadgets weren’t going to burn down homes and so they might write fireplace insurance coverage. Is that also the premise of the corporate? How does that work?

I prefer to say that the premise of our firm was to handle the security of the know-how of the day. And on the time, 1894, World’s Honest, on the sting of College of Chicago, the place each you and I’ve a little bit of historical past, the Underwriters’ Electrical Bureau introduced our founder to Chicago to assist do some main scientific analysis on the security of electrical energy to jot down requirements about how that electrical energy needs to be used, each manufactured and embedded into merchandise and put in and safely utilized in buildings, after which carry out public advocacy, educating folks on the brand new applied sciences.

Quick ahead, definitely with the electrification of the whole lot, the power transition, AI knowledge facilities, electrical energy and electrical security proceed to be a main fear and a driving pressure. However there’s a number of different new applied sciences of the day that we proceed to assist preserve our prospects secure.

What are the opposite applied sciences that you just’re principally targeted on?

Among the most present ones are AI security, the methods wherein AI is being embedded in merchandise, and the methods wherein people have interaction with the security of AI and merchandise. Our latest define of investigation, which is a precursor to writing requirements, was printed in November, and it’s throughout the secure use of AI embedded in merchandise.

That seems like a really meaty topic of dialog right here. There’s a whole lot of AI security debate in our nation and on the planet, so I wish to come again to that.

I simply wish to begin with some foundational questions that I’ve. One among them is the place the authority to inform the business what to do comes from anymore. While you had a bunch of insurance coverage corporations saying, “We gained’t pay your insurance coverage claims if the factor that burned down your home wasn’t UL-certified,” that offered an terrible lot of incentive for folks to go get that testing finished to pay for it. On the time, UL was a nonprofit. Lots of that’s modified since then, proper? The place does the authority or the inducement to take part within the UL course of come from anymore?

It’s a extremely essential query and relevance is a extremely essential strategic idea that we give attention to lots. Who does it matter to in case your product has been licensed to a UL normal and even one other normal? We certify to over 4,000 requirements. Just one,500 of these are precise UL requirements. There are different authorities with jurisdiction and requirements improvement organizations globally.

The significance of that is that governments and definitely insurance coverage corporations, underwriters, even at the moment — and within the US tort system that turns into essential— want to be certain that what they’re underwriting is secure, what varied companies of governments around the globe deem secure. How do you proceed to construct that belief between shoppers and companies and be certain that folks consider that the merchandise that they’re utilizing are as secure because the requirements permit them to be?

That seems like a reasonably large combine. You continue to have insurance coverage corporations saying, “You want UL-listed gadgets in your home, or possibly we gained’t pay claims,” or, in the USA specifically, if there’s litigation across the security of the merchandise, this certification goes to be essential. You may need some governments insisting on varied logos. I believe we are able to all see the certification logos on the merchandise we’ve.

Is it a combination? How do you because the CEO take into consideration, “Okay, these are the constituents who need this emblem. I’m going to go take their wants and inform the business, notably the tech business, which doesn’t prefer to hearken to anybody, that they need to take part.”

Oh, they don’t. I began my profession there.

How does that dialog go?

It goes like this, and I’m going to provide a extremely nice instance. Let’s speak about e-bikes, and specifically e-mobility gadgets, however e-bikes in New York Metropolis. About 5 years in the past there have been a pair dozen individuals who had been killed in New York Metropolis, and why? Overcharging of the lithium-ion batteries.

Lithium-ion batteries have a distinct chemical composition. The thermal runaway occurs quicker. The chemical substances are tougher to place out. In a typical home fireplace, you have got a pair minutes to get out. With a lithium-ion battery fireplace, you have got fewer than 30 seconds to return out alive.

So that you’ve acquired this drawback. Individuals are dying. You’ve acquired this different drawback, which is individuals are excited to make use of e-bikes as a result of they’re an inexpensive mode of transportation. They’re a really helpful merchandise. So how do you stability this?

We at UL Options heard from a variety of prospects, labored with our not-for-profit companion, who’s our largest shareholder, UL Requirements & Engagement, to jot down three requirements across the secure charging, the usage of batteries, and the methods wherein lithium-ion batteries had been put in in e-bikes. Three requirements. We went to New York Metropolis, labored with the mayor’s group and the fireplace providers group there, to make sure that these requirements had been written into New York legislation.

As soon as an ordinary is written into native laws, if you happen to’re a bicycle producer, you’re not going to fabricate a distinct bike or a distinct charger to promote into New York Metropolis than you’d in Chicago or Toronto or LA. So it begins to proliferate.

The excellent news is that since these requirements had been adopted in New York Metropolis, deaths have dropped by 75%. There’s a actual want for the security of humanity in these requirements, after which that turns into picked up by different authorities having jurisdiction, different communities like these different cities I named, and even native personal campuses. Universities have expressed curiosity in, “What are these requirements? How can we take into consideration guaranteeing {that a} dorm doesn’t catch on fireplace?”

That’s the genuine method to how this occurs. There must be the security science that exhibits what the reply may very well be and needs to be. After which there must be a recognition that that want is actual and that it helps promote that belief between these authorities having jurisdiction, these governmental our bodies, and the residents and the customers of merchandise inside their jurisdiction.

It’s fascinating as a result of the choke level there may be retail, proper? Town just isn’t going to allow you to promote a motorbike with out the certification as a result of it’s deemed the bikes with out the certifications to be harmful. Is that constantly the sort of incentive that makes folks undertake the requirements or the certifications, that you need to cease, that there’s enforcement someplace?

Not all the time. We’re going to speak extra about this AI normal, UL 3115, however that began with our prospects coming to us. We see this lots, our prospects saying, “Hey, as a producer,if there’s a normal that we should always undertake and that we all know our opponents will undertake, that ranges the taking part in discipline and creates a constant market.” I spent virtually 20 years in manufacturing. Our prospects steadily come to us and say, “We see this taking place. Assist us take into consideration how this new innovation, this new know-how ought to contemplate what the security science is.” That turns into the precursor to writing an ordinary.

Regularly our prospects don’t even watch for the usual to be written. They begin utilizing that define of investigation to information their product design and innovation in order that they’re extra assured coming again to that insurance coverage query, that if one thing occurs they won’t have a failure in security.

I wish to come again to the notion of consumers, as a result of UL has been restructured because you’ve been there. You took UL Options public. I’m very interested in that set of incentives and what which means.

Each time I discuss to any individual who runs a requirements group, and we discuss to lots right here at The Verge — whether or not it’s Bluetooth or HDMI — there’s just a few factor of being a politician that’s concerned in that. You wouldn’t consider Bluetooth as a deeply political group, however they’ve a whole lot of unwieldy stakeholders who’re pulling in several instructions. You had been describing it as, “we have to create a market.” With HDMI, possibly you desire a characteristic that nobody else needs, and that’s a political drawback for that normal. It doesn’t look like you have got that very same set of pressures. How a lot politicking do you do?

We actually don’t do politicking. Within the requirements improvement course of, it’s a consensus course of. As I stated, our prospects steadily come to us with the necessity for the standard. AI knowledge facilities are an ideal instance. Shifting to 800-volt DC is a really important power want and security problem. How can we begin constructing requirements round that? We kick that over to UL Requirements & Engagement, who’s really the requirements improvement group, the place they convene technical panels and observe a consensus-based course of. There’s some fairly rigorous approaches to that normal improvement and the consensus grounded in science.

Now getting that normal adopted by governments does take… And once more, our requirements improvement group does this, the not-for-profit. They’re concerned in guaranteeing that the proper consideration is given to the chance to undertake these requirements and spend their time selling why it’s an essential want, why it’s a good suggestion.

Let me ask about this construction then, since you are describing the interior relationship between the three components of UL. It began off as clearly one large group. It’s now been reorganized into three subsidiaries. Why the change and what are the divisions right here?

We had been not-for-profit from our founding in 1894 till 2012. We had been based to do the security science analysis, the requirements improvement, and the general public advocacy. Instantly following the World’s Honest, corporations began coming to Underwriters’ Laboratories asking for his or her merchandise to be examined, inspected, licensed. We did that as a not-for-profit, however charged for that, till 2012.

In 2012, our trustees realized that our opponents, many European, who had been based with comparable histories as not-for-profits, had the chance to each do a greater job funding the not-for-profit facet and unleash that for-profit power in an more and more aggressive surroundings. So in 2012, we break up the 2.

I joined in 2019 as CEO of the for-profit with the relationships again to the not-for-profit across the requirements improvement and the analysis. Immediately they’re structured as three separate entities. The requirements improvement group is the shareholder of UL Options. Once we went public in 2024, two years in the past final weekend, it was a secondary providing and so they obtained the total set of proceeds to fund their endowment for his or her requirements improvement and analysis institutes. So we’ve acquired a separate board of trustees, and 4 trustees sit on our board of administrators. So there’s a good strategic relationship, and I believe that that’s essential, however we’re run fully individually.

So there’s the three organizations: UL Requirements & Engagement, the UL Analysis Institutes, and UL Options, which you’re the CEO of. Options is a public firm, however you’ve acquired the trustees of the nonprofit in your board. How a lot do they get to inform you what to do?

I used to be a public firm CEO previous to becoming a member of UL Options, and I don’t see any distinction between this board and my earlier board, as a result of there’s a distinction. My earlier board, I had Berkshire Hathaway as my largest shareholder, and they didn’t sit on our board. I used to be well-trained that as CEO, because the administration group, and because the board, we serve all shareholders, not a single shareholder.

I deal with that in the identical vein right here. All of our shareholders deserve equal consideration and responsibility of care, responsibility of loyalty to all of them.There’s strategic worth in having the proper strategic relationship with the not-for-profit, and that worth goes to all shareholders. That’s the way in which we give it some thought. That’s the way in which we deal with our board conferences. That’s the way in which we deal with our administration choices.

I’m very curious in regards to the business incentives you have got operating the for-profit a part of the group. I perceive you had a whole lot of opponents that turned for-profit testing labs, and I do know the Decoder viewers sufficient to say, “Properly, that clearly corrupted them. They’re simply promoting marks now and promoting extra marks makes them extra money and possibly the testing requirements have gone down.” And I’m curious the way you stability that.

I hear that from our viewers lots, that the financialization of the whole lot has corrupted the whole lot and the belief is gone as a result of everybody’s simply chasing {dollars}. You run a public firm, you’ve acquired shareholders, you’re over right here speaking about them. How do you handle that? You in all probability might reduce the requirements and promote extra certifications, and that may in all probability be higher to your shareholders within the quick time period, however clearly there’s the long run of the model and what it means to folks and the nonprofit to guard. How are you balancing all of it?

We’ve been round for 132 years and we nonetheless communicate the phrases of our founder, that are, “Know by check. State the information.” If we had been to ever deviate from the very best high quality requirements, if we had been ever to deviate from the very best high quality science, it might erode the belief that our prospects have in us that we’ve constructed for 132 years, and our enterprise is belief. I fervently consider that we’ve to proceed this long-term view of progress and relevance: develop so far as our affect and our skill to advise our prospects and help them, however stay related. Tthe solely means you stay related is if you happen to preserve that belief.

While you say prospects, you don’t imply shoppers, proper? You don’t imply the tip consumer. You imply large corporations, governments. How do these prospects specific their preferences to you available in the market?

We’ve three segments to our enterprise: industrial, the place prospects are usually promoting their product within the B2B house; shopper, the place our prospects are usually promoting their merchandise into B2C house; after which our threat and compliance software program section, the place these are usually our largest multinational, world, and strategic accounts.

Our groups are on the market working with the brand new product improvement groups, the standard and compliance groups, in all of our prospects. And our prospects specific their wants. As they’re going by their innovation cycles, we steadily have a line of sight into their product street maps and the way they intend to make use of know-how otherwise in innovation.

I say steadily, “innovation with out security is failure,” and I believe our prospects really feel that very same weight. They don’t wish to fail. They don’t wish to have a product launch that’s going to hurt any individual both in that industrial surroundings or that shopper surroundings.

It’s a extremely open, trustworthy dialogue as a result of we’re there to assist them. Typically serving to them is giving them information that they don’t wish to hear. Nevertheless it’s incumbent upon us to inform them, “These are the information, that is what occurred within the check, and now you need to return and do one thing about it.” We will’t advise them on the right way to redesign their product. That might be a breach of that belief. We’ve to remain agnostic and check when that product pattern is available in.

I’m very interested in that. You stated prospects don’t wish to make merchandise that harm folks. The tech business says that to us lots. And specifically in AI, they are saying this to us. They speak about alignment and security on a regular basis, after which we are able to all see the reporting about what chatbots are doing to shoppers. The place is that stability? Is all of it simply industrial purposes? We don’t need the AI to run the elevators fallacious? Or are you wanting all the way in which to the mannequin capabilities?

We give attention to merchandise. We give attention to product security.Useful security of merchandise could be once you embed software program, let’s say, in an electrical automobile, you don’t wish to flip the radio on and have the brakes slam as a result of the most recent software program obtain modified the if-then-else statements and you end up in a security drawback. Equally, with AI, you wish to guarantee that AI just isn’t creating useful security challenges. And we’re listening to from our prospects that additionally they wish to be certain that they will profess belief within the fashions.

Our UL 3115 got here from prospects coming to us and saying, as an ideal instance, a toddler’s toy. How are you aware that the information that was used to coach that AI that’s embedded in a toddler’s toy was honest, that it stays personal, that it’s clear, that there’s lack of bias within the algorithm? As a result of all of that determines how that product really performs, and in order that’s the angle that we’ve.

However again to your first remark in regards to the know-how business being very immune to others setting requirements or pointers or rules, we fervently consider that third-party unbiased testing inspection certification results in higher outcomes for society.

I imply, I can level you proper now to AI-powered kids’s toys which might be completely off the rails.

And I’ll simply carry that again to, what’s the enforcement mechanism? What’s the choke level? There’s no New York Metropolis that’s going to say, “You’ll be able to’t promote teddy bears in our city except the AI has a certification.” I don’t assume that exists in a few of these markets. The place are you discovering the enforcement or the inducement construction that makes them take part?

It’s early days, and I fully agree with you that we’ve acquired to get our arms round this. There are a variety of requirements improvement organizations around the globe, not simply UL, however IEC, ISO, others which might be coming collectively and saying that that is essential, that is essential. We are going to proceed to advocate that varied governments and authorities having jurisdiction, tech business associations, and others proceed to pursue this.

However it’s certainly, an uphill battle the place the tech business likes to have their very own method and can cloak themselves in mental property and proprietary requirements. And I get that. I began my profession at IBM, I perceive the worth of tech and IP. However I’m a lifelong security freak and I actually consider that some of these things might make merchandise nherently unsafe, and we have to do our greatest to stop that from taking place.

Let me ask you the opposite Decoder query I ask everyone, after which I wish to dive into it utilizing that framework. How do you make choices? What’s your framework for making choices?

My private framework is grounded in knowledge. I’m a knowledge individual and I believe that you must have sufficient knowledge and stress check it to make resolution. I consider organizationally in empowering folks; in case your job is to run X, then you have to be grounding in knowledge and making choices round that, after which there’s a sure degree of selections that ought to probably get bubbled as much as me. However a whole lot of occasions, I believe the folks closest to the client, closest to the selections must make that.

There’s one set of selections right here at UL that I’ll by no means overrule, and that’s the scientific choices that our scientists, our engineers, our lab technicians make. Each occasionally, a buyer just isn’t proud of a report or a choice that we’ve made and it might get raised to me, and I believe my group has the boldness to know that I’ll by no means overrule a scientific or engineering resolution.

That appears essential. That’s the center of the enterprise, to guard the sanctity of the testing.

Within the context of AI, however even within the context of batteries, which I wish to speak about at size really, it feels just like the market is getting farther and farther away from desirous to comply.I’ll provide the instance right here. The Biden administration actually pushed for AI security and so they had a set of requirements that they needed to promulgate. President Obama was on the show talking about the necessity for AI security. And his comparability was, very explicitly, “We failed to control social media and harm folks, we’re not going to screw that up with AI. We wish the labs to publish their testing on the very least.” Trump administration confirmed up, all that’s mainly gone. That Biden-era EO is not in impact, it’s a free-for-all. What’s bringing the Frontier Labs to the desk with you? What’s bringing OpenAI or Anthropic or xAI to the desk?

I’m optimistic that there are world forces round this. As a result of, once more, multinational corporations don’t simply must observe regulation in the USA. They should observe what’s taking place within the EU, what’s taking place throughout Asia. And once you have a look at the affect of various international locations and totally different authorities having jurisdiction in a few of these subjects, I do assume it’s going to develop. However I agree with you, there’s not any sort of top-down-forcing perform proper now to carry them to the desk.

Are you engaged with OpenAI or Anthropic or xAI or Meta?

We’re in a roundabout way engaged with OpenAI or Anthropic. We definitely have finished a good quantity of labor with Meta by the years and a lot of the hyperscalers and extra on the product facet. However these proceed to be subjects of dialog that our chief scientist and our PhD researchers in AI are on the market selling and persevering with to attempt to push the rock up the hill.

You talked about your new normal, UL 3115. It’s a reasonably wide-ranging normal, proper? It’s the whole lot from knowledge facilities to shopper purposes. I believe the primary two merchandise licensed beneath it are out, or the certifications had been obtained and so they’re constructing management purposes, from what I perceive.

Yeah. That was the Hanwha Qcells announcement.

That to me is, “Okay, we’re going to certify a constructing management utility to verify it doesn’t go haywire and switch up the warmth in all of the models,” or no matter a constructing management utility would possibly be capable to do. All of the elevators are going to go loopy.

That is only a philosophical query. These AI programs are essentially nondeterministic. They’re not predictable in the way in which that they function, and that truly is what makes them highly effective. There’s the dangerous facet of hallucinations and them posting to their very own bizarre inside Fb that they’ve constructed for themselves. After which there’s the nice facet of, oh, which means they’re inventive. They will do software program improvement in a means {that a} deterministic system actually couldn’t do earlier than. How do you check that? What’s the mechanism of testing whether or not an AI-powered constructing management software program is all the time going to do what it says if the engine powering it’s inherently unpredictable?

AI fashions actually relaxation upon that predictive modeling, however our focus just isn’t on stepping into the black field of the code. Our focus is on establishing over 200 standards round how, internally, after they’re making choices about their code improvement, they need to take into consideration bias, how they need to take into consideration transparency, how they need to take into consideration equity and privateness.

While you say “assume,” is it the fashions considering or is it the folks making the fashions considering?

The folks designing these fashions. How are they constructing out, what’s the veracity of the information supply that they use to coach the fashions? That’s outlined in our normal of how they need to make choices. I like that you just give attention to how choices are made. Once I have a look at UL 3115, I believe that it’s a normal to assist information these choices as AI is being embedded in merchandise.

The massive alternative in AI proper now could be software program improvement. The price of producing new software program is dropping precipitously and will drop to zero as a result of the instruments are so good at it, and instruments like Claude Cowork and OpenClaw can simply go do issues for you on a regular basis, which is absolutely fascinating.

Meaning the variety of suppliers of AI-empowered software program is simply going to skyrocket. While you describe the market-making functionality or the market-making perform of UL, that “everybody goes to get this certification so we’re all on the identical degree taking part in discipline,” if the taking part in discipline is huge and it’s a bunch of youngsters writing purposes of their basements who don’t care about you, it’d simply completely get away from you. How do you consider that stability of huge gamers who wish to take part and get the emblem mark versus an entrepreneur saying, “I could make you this constructing management software program less expensive,” who by no means really involves you?

I believe that’s the place our prospects and what they’re searching for are available, and the way they’re going to degree the taking part in discipline of their competitors. Sooner or later, the tip shopper does communicate. I used to be in manufacturing for 20 years. I don’t need unsafe AI-powered kilns or metallic presses in my surroundings. There’s some extent at which you’re going to need that verification, that validation, that endorsement, that what you’re putting in in your industrial surroundings or what you’re bringing into your house as a shopper is secure.

That’s the place I do assume the tip consumer has a voice, as a result of they’re going to resolve, “Do I wish to purchase this product or not?” We’ve loads of exams we try this have completely nothing to do with an precise regulation, however need to do with the truth that our buyer has determined that that is essential for his or her model, for his or her finish shoppers, and that drives the demand for what we’ve to supply.

The opposite dynamic that’s taking place in AI particularly is that the fashions themselves are getting ever extra succesful and the concept that you must construct a particular AI utility that’s a wrapper across the mannequin that forces it to do what you need, who is aware of how that’s going to play out. However you possibly can see, “Properly, possibly really I simply want a subscription to Claude and I don’t want a subscription to some utility that’s powered by Claude as a result of Claude can simply do it for me.” If these corporations aren’t engaged with you, how does this work? If Anthropic and OpenAI and the remaining aren’t engaged with you, how does this work?

It’s a extremely nice query as a result of one of many considerations or questions that I even have about AI comes again to that veracity of the coaching knowledge. Again in my coding days, it was “rubbish in, rubbish out” — the extra rubbish that will get in to coach these fashions, the extra issue you have got trusting that these fashions even have the efficacy into the long run and gained’t simply spiral upon themselves and grow to be ineffective. I consider that ought to actually be interesting to those improvement corporations round, “Does that mannequin have the longevity to proceed to supply the solutions, the intelligence, the data that’s grounded in one thing that’s really true and proper?”

I’m simply going to ask you straight up. Do you assume they care?

I hope they care, as a result of it needs to be self-preservation for them to care.

I imply, they appear to be doing fairly properly with out caring. That’s why I’m asking.

Properly, there’s quick time period and long run, however we’ll see how this performs out.

You talked about the stress to rein in, be extra secure, have extra management for AI would possibly come from different governments, different organizations. Possibly it’s the states. The place do you see probably the most stress on making AI secure come from proper now?

It’s fascinating, and I do know you’ve spoken with a few of our massive prospects not too long ago.I believe it’s coming from these massive multinational world prospects who care deeply about how their merchandise are utilized in environments and need their relevance and longevity to be on the market. They don’t wish to discover themselves in a state of affairs of failure.

While you speak about these prospects, are they coming to you and saying, for one thing like UL 3115, “That is what we’d like it to say in order that when it exams it meets our wants”? Is that how that normal is developed?

No, they’ve come to us and stated, “We want an ordinary, assist us give it some thought.” And in order we begin to develop it, we carry them right into a room after which we’ve acquired our PhD AI researchers in there with them.It’s a dialogue grounded in science with then a consensus of, “Okay, we expect that this really will actually assist us. Let’s guarantee that that’s in there.”

PhD AI researchers are very costly these days.

Are you able to pay on the high of the marketplace for these of us?

We’ve constructed a small and mighty group on this and we really feel superb about their thought management and what they’ve contributed.

I’m curious as a result of that’s the opposite arms race. I have a look at this from the surface and I say, “nobody can sustain with these labs. They’re paying all the cash. Even the competitors between them doesn’t appear to be holding them in examine.” The concept they’re all going to enroll in a literal checkmark from UL that claims they’re secure after they’re all racing to an IPO… I’m simply very curious the place that stress goes to return from. I don’t know if it’s going to return from an industrial manufacturing provider at this time limit. I believe it may need to return from a authorities.

We’ve acquired to maintain pushing this rock up the hill. It’s nonetheless early days and it’s essential to determine it out.

The opposite piece of this, as you talked about, the usual covers knowledge facilities. There’s a whole lot of pressure, political and in any other case, round knowledge heart build-out on this nation and in all places else. There’s simply {the electrical} part of it, proper? If you happen to’re going to do a whole lot of electrical energy in a room, you in all probability desire a bunch of UL-certified parts in there. Is there greater than that in UL 3115 because it pertains to AI knowledge facilities?

UL 3115 is simply actually round AI embedded in merchandise. With AI knowledge facilities, there’s 70 different requirements that we check to at the moment across the security of the electrical energy, the parts, the chillers, the DC present coming in, the inverters, all of that. Then there’s an entire host that we’re listening to from our prospects that with the speedy change within the quantity of energy, the speedy change simply within the thermal dynamics of GPUs versus CPUs, the speedy change in the way in which that you just’re going to place a megawatt of energy right into a rack or that you just’re shifting the water cooling. There’s an entire set of recent requirements means outdoors of UL 3115. We’ve had two AI knowledge heart summits with prospects on how they’re enthusiastic about their wants for requirements in knowledge facilities and the way we are able to quickly assist them proceed to develop on their innovation tempo in ways in which they will really feel comfy will likely be secure sooner or later.

Do you assume they’re going to decelerate their build-out objectives to ensure that these certifications to take maintain?

No, I believe they’re anticipating everyone else to choose up the tempo.

Let me ask you in regards to the different race situation, as a result of once more, I believe it might be nice if the whole lot was licensed and the whole lot was secure. After which I have a look at the markets that we’re in and there’s simply an explosion of issues on a regular basis. The one that basically strikes me is the whole lot with a battery in it. We’re profiling increasingly more of those corporations right here at The Verge on a regular basis. If you happen to’ve acquired a lithium-ion battery and a high-efficiency motor and a dream, you can begin an organization that makes 500 merchandise at the moment. We’ve profiled a few of them. Hoto and Fanttik are two which have simply sprung up, and so they make instruments. And the opposite day I noticed a kind of corporations had like a lithium-ion handheld Sawzall, which is simply a whole lot of energy. If you happen to’re going to place that a lot torque in a bit of motor, that’s a whole lot of energy you’re going to attract.

I have a look at these corporations and so they’re clearly all based mostly in China, and whether or not or not they’ve UL certification is irrelevant to the shoppers shopping for all these merchandise.As a result of they’re legitimately cool merchandise and there’s a race of innovation taking place there and it’s all simply on Amazon, and Amazon doesn’t appear to be imposing any of those requirements in any respect. How do you consider that? How do you consider the prevalence of high-powered lithium-ion batteries in all places with out the patron demand to your certification?

To start with, Amazon is a good buyer of ours and you may drop down and see if one thing’s been UL licensed.

They need to make that extra outstanding. I believe you need to in all probability inform them to make that extra outstanding.

They’re an ideal buyer of ours. And certainly, innovation is quick. Batteries are thrilling and harmful, and we proceed to work with customs brokers, varied authorities having jurisdiction, and our prospects to assist educate the right way to preserve these lithium-ion batteries secure, notably if you happen to’re importing into the US markets. An incredible instance of this was a couple of decade in the past when hoverboards had been exploding and—

Sure. So the Client Product Security Fee got here to UL Requirements & Engagement — UL Options, on the time — and stated, “Are you able to very quickly write an ordinary and assist us get our arms round this?” And we did that. And once more, it helped with the security.

One of many key areas that we’ve is market surveillance and anti-counterfeiting. So we’re consistently working with customs brokers and likewise with opponents who’re placing the UL mark correctly on their product, who will spotlight a product that’s available in the market that’s not assembly the codes and the requirements. We’ve gained some important lawsuits round these circumstances the place there are unsafe batteries, uncertified conditions after they’re not in compliance with the legislation.

Amazon and UL collectively, you’re suing some e-bike manufacturers which might be promoting on that platform with faux UL certifications.

It’s a must to catch them. So you have got an enforcement group that’s really scanning Amazon for faux UL certifications?

We’ve a group that works with anybody who needs to spotlight that they assume that there’s a unsafe state of affairs or a counterfeit UL mark, our group responds.

Are you able to scale up quick sufficient to fulfill the flood of recent merchandise? Once more, with a lithium-ion battery, a high-efficiency motor and a dream, you can begin an organization and make 500 merchandise tomorrow. Are you able to scale as much as meet that flood by way of testing?

We will completely scale. We’ve scaled everywhere in the world, and we prefer to say we meet our prospects the place they’re. If you happen to’re doing innovation in China, we’ve acquired our testing labs in China able to go. If you happen to’re doing innovation right here in the USA, we’ve acquired our labs right here able to go. If you happen to’re manufacturing anyplace on the planet, our discipline inspection group will go to your plant 4 occasions a 12 months to make sure that you’re manufacturing in accordance with the requirements that we examined to. We’ve been rising, and we’ll proceed to develop.

Do you make the case to Chinese language producers, “Hey, you probably have this UL certification, you’ll make extra gross sales”?

And there’s knowledge displaying that the US shoppers really care about this?

Sure. And producers in China, all throughout Asia, they know that in the event that they wish to get their product into the US market, they should observe the security requirements and we’re there to certify for them.

I do know you just made an acquisition to develop your testing presence within the EU. How large is your presence in China in comparison with the USA and the EU?

We report income by level of buyer. If you happen to’re a US buyer however we’re testing your product in China since you occur to have an innovation heart there, we’ll report that within the US. Final 12 months, I consider 42% of our income was level of buyer in the USA, 25% of our level of buyer is China, about 17% is EMEA, after which the remainder of the world. So China has been essential for us. We’ve been in there for 40 years. We’ve acquired a three way partnership partnership and we’ve independently wholly owned labs as properly. We work very intently with a big variety of Chinese language producers to assist them get their merchandise to markets everywhere in the world.

The connection in China has been the purpose of competition not too long ago with the Trump administration.

In the course of the Biden administration, the FCC launched something referred to as the Cyber Belief Mark, which was purported to certify IOT gadgets particularly as being secure. UL was purported to be the lead administrator form of writing the usual for that.

Brendan Carr, who’s well-known to listeners of the present and my different present, The Vergecast, is the present chair of the FCC. He has a whole lot of concepts, Mr. Carr, and he decided that your relationship with and your work in China someway was corrupting. One thing occurred, which I’m dying to know what precisely occurred, UL is no longer participating there and a Trump donor’s firm is now the lead administrator. What was that dialog with the Brendan Carr FCC like across the Cyber Belief Mark?

We’re a proud American firm. We’ve been right here for 132 years. If our authorities asks us to serve, we after all will step up and serve and help no matter they want. And so we had been actually happy with the work that we had been in a position to do as a lead administrator to assist arrange the parameters of that and work intently with the FCC.

When the FCC determined that they needed extra necessities from the lead administrator, we realized that we weren’t the most effective match for that. And we simply transferred that mental property and that work again to the FCC and so they continued down their path.

What had been the extra necessities?

These necessities had been actually round how they needed to run this system sooner or later. And it was a set of necessities that we didn’t really feel that we had been the most effective members to do.

That sounds very bureaucratic and administrative. I’m taking a look at Brendan Carr, he mainly accused you of being beholden to the Chinese language authorities. Did you ever reply to that instantly? How would you reply to that now?

We’ve been very clear about our operations, {our relationships} everywhere in the world, and we proceed to be so.

Brendan is not a subtle man. He doesn’t do things in the shadows. He says you’re beholden to the Chinese language authorities and also you’re saying that’s completely not true. And it was sufficient so that you can stroll away and say, “We don’t wish to be part of this.”

I believe that the place all of us landed is the proper reply for all of us.

Equally, the FCC proper now could be banning a bunch of Wi-Fi routers just because they’re made abroad. Clearly, you regarded into this with the Cyber Belief program, you have got these different certifications. Do you assume it’s appropriate to say any gadget made in China is an inherent safety threat?

We’ve lengthy and deep relationships with prospects everywhere in the world and lengthy and deep relationships with prospects in China. These prospects see worth in testing to requirements and following rules and guidelines, and we’ll proceed to help them within the ways in which they want.

Do you assume that there’s a possible certification for gadgets made abroad that US shoppers or US corporations can say, “Okay, the provision chain threat that we’ve heard about has really been mitigated or the suitable controls are in place”?

I believe the set of requirements that exists at the moment actually facilitates that belief that buyers ought to have with merchandise which might be made anyplace on the planet. If you happen to’re adhering to these requirements, if you happen to’ve acquired a third-party tester that has endorsed and licensed that you just’ve met that, I believe that’s the mechanism that does that.

I simply see the proliferation of merchandise and I’m questioning if possibly all the way in which on the finish, you say you have got some knowledge that claims shoppers want UL merchandise and I hope that’s true. However then we at The Verge cowl, I don’t know, cameras for your home which have simply gaping safety holes in them, the place there’s identical to stay feeds streaming to the entire web at massive as a result of there isn’t a safety equipment or an updates equipment. We do see that with routers. We’ve seen a whole lot of hacks with consumer-grade routers. I’m simply questioning the place that extends to, notably in software program.

You purchase an influence strip, you possibly can see the emblem on the again of it, or possibly Amazon will not less than present you the emblem and possibly you’ll nonetheless purchase the cheaper one since you don’t actually know what it’s for. With these software program merchandise or these {hardware} merchandise which might be operating a whole lot of software program, it’s not proper in entrance of you. So how do you make that case?

We do have a service that focuses again to that useful security of that embedded software program or that efficacy of that product being linked to the web and its cybersecurity. There are requirements round that and there are methods to method it, however I believe what you’re highlighting is a chance to make shoppers extra conscious of what they need to be searching for and demanding as they buy their merchandise.

Do you assume that that is only a market drawback? I believe possibly that is what I’m coming again to time and again all through this dialog. I actually want the patron market demanded extra of those corporations. However that’s only a collective motion drawback.

I believe it’s completely rational for most individuals to simply choose the most affordable energy strip that Amazon has on the primary web page, and I can’t actually blame them for it. On the identical time, possibly we don’t have a federal regulator who’s going to step in and say, “Okay, to maintain everyone secure, we’re going to demand the certification.” Possibly we don’t have insurance coverage corporations who’re going to go demand them of Amazon.

After which relating to software program, it looks like the tech business specifically is totally immune to anybody telling them what they will do. And the concept of a UL certification for firmware updates in your cameras on a cadence is simply possibly the toughest promote of all. So if it’s not going to be the shoppers that do it and we’ve a authorities that appears checked out of it, that is what I preserve circling and what I used to be most excited to speak to you about. The place does the stress come from for folks to take part in a security program?

That is, to me, one of many thrilling items of after we went public and funded the endowment for our not-for-profit. We’ve talked lots in regards to the requirements improvement group, UL Requirements & Engagement. We haven’t talked a lot in regards to the UL Analysis Institutes and the areas the place they’re targeted.

One among their institutes is concentrated actually on AI security and the way ought to the world be higher educated on what could be thought-about secure and the place they need to dig deeper. There’s much more to return, not simply on the analysis round that, but in addition across the step to boost the patron’s consciousness of the truth that, if one thing’s free, you’re the product. Again to social media, if you happen to’re utilizing it and it’s free, you’re the product. How do you defend shoppers from that? It’s a extremely essential idea and I nonetheless assume it’s early days on this in AI.

You’re within the enterprise of promoting security. I believe that’s a good solution to describe what UL does. Do you assume that the way in which that Dario Amodei or Sam Altman speak about AI alignment and security is efficient? As a result of their pitch is, “If you happen to don’t allow us to do no matter we wish, we would kill everybody on the planet.”

I believe they’re making an attempt to floor in science and engineering and definitely in several methods to make use of AI and totally different fashions. LLMs are one method, however there’s a number of others. It’s in all probability a false option to say, “Allow us to do what we wish and subsequently we’ll forestall this from destroying.” I believe you want each.

While you say each, you imply outdoors testing and validation or authorities regulation? What do you imply by that?

The entire above. It might be splendid if the tech corporations got here collectively and stated, “Right here’s what we consider collectively will assist preserve the world secure after which we’ll adhere to that,” versus letting every one simply go off and observe no matter path they assume is greatest.

Once more, you handle a sophisticated security construction, so I’m simply asking you abstractly. If you happen to needed to choose a construction for that to occur in, does that seem like a authorities regulation? Does it seem like an business physique? Does it seem like a nonprofit that controls a for-profit testing heart? How would you design this?

I believe the place it’s heading is towards extra of the requirements improvement organizations and the business our bodies coming collectively, as a result of they would be the most educated about what ought to work. You all the time need that deep business experience once you’re growing any kind of security normal that then strikes into regulation. If you happen to begin with regulation top-down, you don’t all the time get to the proper reply and it’s not all the time grounded within the science and the engineering that it must be. I’d advocate business teams with requirements improvement organizations.

Jennifer, what’s subsequent for UL? What ought to folks be looking for?

For us, it’s going to be this continuation, as I prefer to say, of progress and relevance. We are going to proceed to be on the forefront of innovation and proceed to seek out methods to make security related for no matter innovation comes subsequent. I can geek out and get enthusiastic about quantum for a second as one thing that’s the long run extension of what’s post-AI or what makes AI higher. However these are areas that we proceed to attempt to keep concerned in and take into consideration — not simply {the electrical} security of 132 years in the past or {the electrical} security wanted in knowledge facilities at the moment, however what’s coming subsequent.

I do like that despite the fact that it’s been a very long time because you’ve been at IBM, you introduced it again to quantum. It’s very IBM of you. I actually admire that.

It’s. I’ve to say, I’ll give a bit of IBM shout-out as a result of I like Arvind (Krishna, IBM CEO).I used to be strolling by O’Hare and so they have their IBM quantum chandelier sitting proper there subsequent to the dinosaur. And I imply, I skidded to a halt whereas I used to be pulling my baggage. I’m like, “Oh my gosh, it’s a quantum chandelier.” It’s actually thrilling as a result of I’m right here in Chicago and we at UL have been concerned in creating the quantum ecosystem that Chicago has been selling and we’re enthusiastic about what’s subsequent in that set of know-how. We will speak about that one other time.

The second somebody ships a working quantum laptop that does economically related duties, we’ll have you ever again to speak about it.

I don’t know when that’s going to be.

Nearer than we expect, I hope.

That’s a daring prediction. Thanks a lot for being on Decoder. This was nice.

Thanks. Good to fulfill you.

_{Questions or feedback? Hit us up at decoder@theverge.com. We actually do learn each e-mail!}