Latest / EU-Startups Podcast / From MI6 to Startups

Transcript

From MI6 to Startups · EU-Startups Podcast

• 0:00
  Speaker 1: You are now listening to the EU startups podcast.
• 0:13
  Speaker 2: And welcome to this episode of The EU Startups Podcast, as always I'm your host, Dade Senangarcia, the newsletter here at EU startups.
• 0:20
  Speaker 2: today i chat with Tyler Edwards.
• 0:22
  Speaker 2: he's the CEO and co-founder of the British cybersecurity startup Overmind.
• 0:28
  Speaker 2: Now Overmind is interesting in comparison two other companies that we've had on this podcast because it just recently raised its seed round back in February.
• 0:38
  Speaker 2: But the real interesting part and what really brings value to The Conversation was Tyler himself, his background in MI-Five and MI-Six helping shape national risk intelligence and AI attitudes within government programs.
• 0:55
  Speaker 2: We were able to chat about development across Europe, the way China and the US are impacting.
• 1:03
  Speaker 2: The innovation sector specifically AI as well as the regulation that AI has within Europe.
• 1:10
  Speaker 2: so I hope you're looking forward to them.
• 1:11
  Speaker 2: but before a quick message from our sponsor.
• 1:14
  Speaker 2: Fanta
• 1:15
  Speaker 0: You're startup founder.
• 1:16
  Speaker 0: finding product market fit is probably your number one priority.
• 1:20
  Speaker 0: But to land bigger customers?
• 1:24
  Speaker 0: and obtaining your SOC-II or ISO-Twenty-Seven-O-O One certification can open those big doors, but they take time & energy pulling you away from building & shipping.
• 1:35
  Speaker 0: That's where Vanta comes in!
• 1:36
  Speaker 0: Vanta is the all-in one compliance solution helping startups like yours get audit ready and build a strong security foundation quickly and painlessly.
• 1:46
  Speaker 1: How?
• 1:47
  Speaker 0: Vanta automates the manual security tasks that slow you down Helping you streamline your audit And The platform connects you with trusted VCSOs to build your program, auditors to get through audits quickly and a marketplace for essentials like pen testing.
• 2:04
  Speaker 0: So whether you're closing first deal or gearing up growth, Vanta makes compliance easy!
• 2:10
  Speaker 0: Join over eight thousand companies including many Y Combinator and Techstars startups who trust Vanta.
• 2:17
  Speaker 0: Simplify Compliance & Get One Thousand Dollars Off at vanta.com.
• 2:21
  Speaker 0: slash EU Startups That's V-A-N-T-A dot com.
• 2:25
  Speaker 0: slash EU startups.
• 2:27
  Speaker 0: Well,
• 2:28
  Speaker 2: hello Tyler and welcome to the Eustarps podcast.
• 2:31
  Speaker 2: it is a pleasure on this podcast to begin with.
• 2:44
  Speaker 2: But that's, I think in this occasion it merits an exception because i find interesting what Overmind does...I don't think we get that many sort of AI cybersecurity people here and you have such a interesting background That is worth even just hearing about for anybody who's tuning right now.
• 3:04
  Speaker 2: So thank-you so much!
• 3:05
  Speaker 2: And before we go into the questions Yeah,
• 3:12
  Speaker 1: of course.
• 3:13
  Speaker 1: And yeah thank you so much for having me on!
• 3:16
  Speaker 1: So I guess the most interesting bit that offers us is usually my background at least to start... ...so i don't come from a typical startup background especially in the UK.. ..so it started my career in defence and intelligence.... ..so did about six seven years with MI-V, MI-VI GCHQ which are three intelligence agencies in the
• 3:35
  Speaker 1: U.K.,
• 3:35
  Speaker 1: those who aren't initiated came in doing offensive cyber.
• 3:39
  Speaker 1: So this was building a bunch of tools and capabilities as we refer to them, all the weird wonderful things that you can imagine the intelligence agencies do before moving on into more sort operational role.
• 3:51
  Speaker 1: And over time I was there managed working pretty much every large global event at the time.
• 3:58
  Speaker 1: so Afghanistan Ukraine or other bits Before spinning out to you found Overmind Betrayal.
• 4:03
  Speaker 1: soon get onto Inner Movement
• 4:06
  Speaker 2: Perfect.
• 4:07
  Speaker 2: Yeah, well before we get into overmind I think i'm interested to kind of hear about what first you drew Into that world because I know also i've seen from your background uh policy advisor in white hall as Well.
• 4:18
  Speaker 2: so just a kind of what pulled You into?
• 4:20
  Speaker 2: That and then separately one moved you into the startup entrepreneur world From it.
• 4:27
  Speaker 1: yeah So i'd always been A massive nerd but basically i'd Always wanted To work in tech.
• 4:32
  Speaker 1: um And It didn't feel like the pool from the larger companies, say Google Meta or the rest.
• 4:39
  Speaker 1: It didn't feel like they were doing things that I necessarily found really interesting.
• 4:42
  Speaker 1: there we're building cool products and attacks obviously impressive But they weren't solving these sort of, like world changing problems at the time when I was looking to start my career.
• 4:51
  Speaker 1: And British intelligence very much wants... They were building tech in these heavily constrained environments having to solve unique problems that you just couldn't use off-the-shelf products for and that felt really exciting.
• 5:02
  Speaker 1: plus the work it sounds pretty sexy from the outside, no intelligence and defences all that stuff.
• 5:07
  Speaker 1: so It is a natural pearl!
• 5:10
  Speaker 1: Then there's the sort of advisory stuff but kind of ended up happening by accident because there wasn't much AI talent in the workforce at that time, not so much more now.
• 5:20
  Speaker 1: I came in towards the end of my career to just help with some of how MOD and government were looking to integrate what... The timers was ML but eventually an AI integrated into intelligence analytics and defence.
• 5:32
  Speaker 1: And it's a very short stint of doing the policy advisory stuff because there was more, more of a technical part.
• 5:38
  Speaker 1: But that is exciting to see how government and MOD intelligence which is famously quite closed environment I'm looking for starts on-board tooling to augment their work.
• 5:49
  Speaker 1: they do.
• 5:50
  Speaker 1: but yeah it's just an incredible place to work.
• 5:53
  Speaker 1: The things i got to do there are really good to do anywhere else some very much shaped sort of how am as a founder now leading this company.
• 6:02
  Speaker 2: Well that shaping of how you are as a founder and without getting to two main specifics I imagine, there's some limitations just the way
• 6:10
  Speaker 1: it is.
• 6:10
  Speaker 1: Yeah
• 6:11
  Speaker 2: but i guess on general level You know Is their moment?
• 6:15
  Speaker 2: You can identify where your working in these risk intelligence communities And come across something say this kind This interesting There was a gap here.
• 6:26
  Speaker 2: Maybe I Can work to fit.
• 6:28
  Speaker 2: yeah
• 6:30
  Speaker 1: so As they said Every single thing that you were faced with was often an entirely novel and unique problem.
• 6:36
  Speaker 1: Either the tech saw it as well, or specific constraints we're building for are just so hostile but have to try find a way in which they can solve these problems.
• 6:45
  Speaker 1: As I started looking at AI adoption this same technical challenge came up where there is no way government would use like American third party tools to start doing like internal intelligence analytics and all these sorts of things.
• 6:59
  Speaker 1: And they especially weren't going to use a third RT generic tool for it, right?
• 7:03
  Speaker 1: These were used cases that are very specific.
• 7:05
  Speaker 1: the date was very specific.
• 7:08
  Speaker 1: so the process of building models that were specialized in these uses fine tune.
• 7:11
  Speaker 1: you know what kind of stuff government just didn't have infrastructure again at the time to do this sort of thing saw very similar happening with commercial space.
• 7:20
  Speaker 1: The trend was general intelligence.
• 7:23
  Speaker 1: biggest, baddest API model was what we're referring at most problems.
• 7:27
  Speaker 1: And to me that just didn't see like the right way for this new technological trends be heading and thus over mine was born.
• 7:36
  Speaker 1: basically
• 7:37
  Speaker 2: okay.
• 7:37
  Speaker 2: Well, you know we're chatting about our mind.
• 7:39
  Speaker 2: Let's get into a overmind here.
• 7:41
  Speaker 2: It it and or minds.
• 7:43
  Speaker 2: interesting too because I barely understand.
• 7:46
  Speaker 2: You know if your nerd And i don't mean that in a bad sense, you Know on the other end of being a nerd which is?
• 7:51
  Speaker 2: Don't understand any of this really.
• 7:53
  Speaker 2: Arts and letters were always my thing but from what I gather it Is an interesting proposal.
• 7:59
  Speaker 2: and you've mentioned hear government use Of programs.
• 8:03
  Speaker 2: is that something that Overmind is kind of shooting towards?
• 8:06
  Speaker 2: cuz you have that in when we're talking about, you know... You understand how governments and intelligence agencies are using programs.
• 8:13
  Speaker 2: Would you be interested in Overmind eventually evolving towards that kind of use?
• 8:17
  Speaker 1: Yeah so I'll tackle both of those.
• 8:19
  Speaker 1: So it's a real big summary of what Overmind does.
• 8:21
  Speaker 1: We have continuous learnings basically with the idea is people should move away from generalized intelligence to models they own specialized their data.
• 8:32
  Speaker 1: So ultimately entering process or collecting data labeling training a model, specialise on that data and then shipping it back out again.
• 8:40
  Speaker 1: And doing the loop-again.
• 8:42
  Speaker 1: The idea is your agent or your model can continuously learn from new data to become increasingly specialised in the thing they're doing.
• 8:50
  Speaker 1: We always say every time you pitch using a frontier model for an agentic task like using a flamethrower to light a candle will get their job done but absolutely overkill!
• 9:01
  Speaker 1: You don't need total knowledge of what we do With AI unless you're doing sort of like quantum physics at the large our drone collider.
• 9:09
  Speaker 1: You don't need a ten trillion parameter model.
• 9:12
  Speaker 1: So yeah, I do as you can ship some smaller more specialized models and that does also work with government and defense.
• 9:17
  Speaker 1: again these are Highly specific use cases whether it's its defense or public health being able to train And hone your model too this specific sort demographic or use case that you're deploying it for.
• 9:29
  Speaker 1: It's far more performative, the model will just be better, it'll be faster and safer than this general API with a system prompt which is never going to be very robust technical setup.
• 9:41
  Speaker 1: so what next?
• 9:43
  Speaker 1: AI has been exciting over these last years year in bit.
• 9:46
  Speaker 1: but What's Next?
• 9:48
  Speaker 1: And for us its owning that stack and specialising intelligence rather then having those general
• 9:53
  Speaker 2: models.
• 9:55
  Speaker 2: Well, ground this for me a little bit because I think i can't be the only one who kind of hears AI risk and just saying okay what does that mean?
• 10:04
  Speaker 2: Is my giant GPT going to hallucinate more you know.
• 10:07
  Speaker 2: Can you give me kind of uh a grounded example of what AI risks could mean in the danger that you see there?
• 10:14
  Speaker 1: Yeah so when we talk about AI risk there's sort of two counts.
• 10:19
  Speaker 1: There is the kind of like catastrophic risk or AIs going to be used to develop bioelectrons, for example.
• 10:25
  Speaker 1: and then this other side a bit of AI hallucinating just not doing things we want not being very performative And it's that slide will be set with not on the sort of category and catastrophic risks I made.
• 10:35
  Speaker 1: because the simplest way to put it is AI models are probabilistic in that they're having general guess at what you've asked them do.
• 10:45
  Speaker 1: And if you want a higher occurrence of the correct guess, You just need to index that little bit more.
• 10:51
  Speaker 1: It's kind like having A bowl of tic-tacs for example and I say there is one hundred in this bowl.
• 10:57
  Speaker 1: Ten of them are poison pills.
• 10:59
  Speaker 1: If i want to reduce the appearance Of poison pills ,I just need To sort of trade them out.
• 11:04
  Speaker 1: So take it few routes, swap those over tic tacs Loosely, that's how the fine-tuning works.
• 11:09
  Speaker 1: We are increasing the probability That the output from model is within that area of speciality you want.
• 11:16
  Speaker 1: And I guess a thing with general big models Is they know everything.
• 11:20
  Speaker 1: They know as much about Shakespeare As they know about melanomas and as they know About A recipe for Victoria sponge.
• 11:27
  Speaker 1: So this sort of probabilistic outcomes You just don't want in your model.
• 11:31
  Speaker 1: If the thing you're doing For example is KYC train your model to do that thing really, really well with a bit specialization in the inside and you're always just going to perform much better because that is the knowledge they focused on.
• 11:43
  Speaker 1: Skipping out of few technical pieces for it-for the sake of time.
• 11:46
  Speaker 1: but if that's the-that's the generogist.
• 11:49
  Speaker 2: Okay okay yeah It's uh cuz.
• 11:51
  Speaker 2: I think a lot people only hear cyber secure start up focusing on cybersecurity even AI Cyber security You know your mind immediately goes into deep fakes or um AI attacks or fishing for your credit card information, that kind of thing.
• 12:04
  Speaker 2: So it's interesting to get a little bit of an insight here into different angle and a different outlook as the way we're progressing with this kind of AI technology.
• 12:15
  Speaker 2: When you raised your seed round which is how I originally heard about you We got your press release was in February.
• 12:23
  Speaker 2: Yes, February this year.
• 12:24
  Speaker 2: Two point three, I think we had euros.
• 12:28
  Speaker 2: And so that in it there was this stat the kind of jumped out to me and It was a Gardner study That was cited said four percent of agentic AI projects could be cancelled due To inadequate risk controls.
• 12:44
  Speaker 2: So i want to hear A little bit about that not specific figures like.
• 12:48
  Speaker 2: i know it's Not yours but what would you say is Kind Of driving The issue to begin with.
• 12:53
  Speaker 1: Yeah, I think...I guess there's two things in that.
• 12:57
  Speaker 1: When we started the company were far more focused on this idea of AI security and we've kind of vacated that space a bit because for us when we're talking to customers about security it was a proxy word for performance.
• 13:09
  Speaker 1: they just wanted you build something did all the thing promised could do.
• 13:14
  Speaker 1: They are willing take some risk with AI third party data access and all the rest, as long as it made them competitively compelling.
• 13:22
  Speaker 1: And they weren't left behind on this thing that their building was performative.
• 13:26
  Speaker 1: And largely The reason we're seeing a lot of these these agentic products being canned is because there just not reaching any of these benchmarks or thresholds.
• 13:33
  Speaker 1: So models aren't performing the way that teams want to be proven speed sort very cumbersome to maintain and develop.
• 13:41
  Speaker 1: technical teams are having to all of a sudden on board either new staff from your skill sets, so control this unwieldy technology.
• 13:48
  Speaker 1: So for the most part people that is defaulting.
• 13:50
  Speaker 1: neither too verticalised tools.
• 13:52
  Speaker 1: they'll just buy something external or use traditional software and use something more deterministic.
• 14:00
  Speaker 1: I think we're kind coming down the other side where we were throwing AI at everything And now we're kind of realizing that, oh okay like as powerful a technology is.
• 14:09
  Speaker 1: We don't need to put AI in absolutely everything and she learning still inches?
• 14:13
  Speaker 1: Yeah yeah well we had the exact same thing with Bluetooth.
• 14:16
  Speaker 1: right all of a sudden.
• 14:17
  Speaker 1: bluetooth was really exciting and popular.
• 14:19
  Speaker 1: iot it's really popular and you can start tweeting from your fridge.
• 14:22
  Speaker 1: how many people can still tweet from their fridge?
• 14:24
  Speaker 1: I think they realized there was not something that that people needed and it wasn't solving our problem.
• 14:28
  Speaker 1: um I think we're doing the same with AI.
• 14:30
  Speaker 1: Everything had AI for a bit.
• 14:32
  Speaker 1: And the kind of calming down on a side bit where just putting AI in something is not enough, we're also trying to work out like do you actually need AI here?
• 14:39
  Speaker 1: Maths still exists.
• 14:41
  Speaker 1: ML's still exist.
• 14:42
  Speaker 1: Deterministic software still exists.
• 14:43
  Speaker 1: Where can we use AI where it's going to be false multiplier?
• 14:47
  Speaker 1: So I think that risk for teams investing all this capital and time into building something was just a complicated piece of tech that isn't actually providing any additional company benefit.
• 14:59
  Speaker 1: At least, that's the area of risk we come from.
• 15:00
  Speaker 1: there are obviously other things relating to data as prevention or all the rest people have focused on.
• 15:05
  Speaker 1: I Have my own opinions and it comes AI but for us It is about how do you help teams utilize this incredibly powerful technology in a way that is enduring?
• 15:15
  Speaker 1: And doesn't mean they're going spend cash time with something?
• 15:17
  Speaker 1: Well,
• 15:19
  Speaker 2: that's in.
• 15:20
  Speaker 2: That's why I love these conversations because it kind of makes me feel less like a neanderthal Because i actually kind of understand what you're talking about.
• 15:27
  Speaker 2: Like this is great.
• 15:28
  Speaker 2: I can follow those conversation sometimes we'll get a press release now read and they will just be gibberish to me outside You know?
• 15:35
  Speaker 2: I don't know What any of this means.
• 15:36
  Speaker 2: And and and that's difficult as a journalist cuz your expected to push these things out there and then you're Kind of left thinking I feel like I need a master's degree too.
• 15:45
  Speaker 2: But
• 15:46
  Speaker 1: so well This Is the thing right In that This this is.
• 15:49
  Speaker 1: it's such a Transformative technology that right now.
• 15:53
  Speaker 1: Is not as accessible?
• 15:54
  Speaker 1: As It should be especially on the granular level.
• 15:57
  Speaker 1: but right Now if you want so, If You decided do you wanted to train a model for yourself On your own data?
• 16:02
  Speaker 1: how would you go about That?
• 16:04
  Speaker 1: You'd have to learn Python, you would get a PhD to learn Tinker and things like that.
• 16:09
  Speaker 1: It's about what can we build it?
• 16:11
  Speaker 1: To bring back bar incredibly low so their teams can do these things.
• 16:14
  Speaker 1: they Can control the technology That's about to define in next few decades of how companies grow and build and all the rest And that's very much our mission statement.
• 16:23
  Speaker 1: there is There's no safe world where this technology is owned just by open AI and entropic.
• 16:29
  Speaker 1: It needs to be something that is readily accessible, buildable by everybody.
• 16:34
  Speaker 2: I think that kind of leads us to a natural topping.
• 16:36
  Speaker 2: This is moving away from Overmind, but you are just already touching on this tension over AI sovereignty because there's different European approaches and the foreign players open AI deep-seek.
• 16:50
  Speaker 2: what have you?
• 16:51
  Speaker 2: How do you see playing out in the next couple years?
• 16:54
  Speaker 2: so much... You've got an understanding at policy level But also get technical understanding.
• 17:01
  Speaker 2: What was your outlook for moving forward?
• 17:04
  Speaker 1: Yeah, I think we're seeing a real resurgence of open source in these last For say year and a bit.
• 17:10
  Speaker 1: I think GTC so that the team was an in San Francisco for a bit.
• 17:13
  Speaker 1: We went to GTC which is NVIDIA's sort big tech conference And So much of the conversation there was about open-waist models.
• 17:20
  Speaker 1: These are models out of Open Source.
• 17:21
  Speaker 1: you can see the internal workings?
• 17:23
  Speaker 1: Um, and i think That Is where The market is starting To move.
• 17:27
  Speaker 1: um, and I Think That Is Sort Of The Natural Market.
• 17:29
  Speaker 1: Pull want technology that they can control, technology they can deploy onto the room sort of infrastructure and all the rest.
• 17:38
  Speaker 1: There is now this massive drive for sovereignty especially in UK Europe to own the tech that has been integrated into again like public health services as well as the open market.
• 17:48
  Speaker 1: so having an entire dependency on one or two US companies it's not a particularly robust approach to things.
• 17:57
  Speaker 1: I think naturally, the tech ecosystem that is pulling us there... There are now what like thirty million models you can download across the ecosystem and its absolutely ridiculous some of them really big in general.
• 18:09
  Speaker 1: Some of them are incredibly specialised and niche.
• 18:12
  Speaker 1: but these natural market forces when they come into tech They're coming here And open source has always been where the innovation really happens.
• 18:20
  Speaker 1: I mean, famously OpenAI started as an open source company before they pulled that ladder up really quickly.
• 18:28
  Speaker 1: so think of it.
• 18:28
  Speaker 1: like we're heading back to is the return of open source and there's very much something that we are trying to purchase.
• 18:33
  Speaker 1: well The power over our product.
• 18:35
  Speaker 1: you can just pick a open rates model And train however your own and then deploy at wherever you want It has yours.
• 18:43
  Speaker 1: So eventually working with sort of government and defence on this AI so that the piece would be interesting.
• 18:50
  Speaker 1: And there again, I have my own opinions in start-ups working with governments especially in the UK.
• 18:54
  Speaker 1: but maybe as a conversation for another side?
• 18:56
  Speaker 2: Sure sure yeah and we do want to keep in mind here it is for entrepreneurs.
• 19:00
  Speaker 2: when i try get over those chat... As interesting Um, but having said that and I am going to slightly go into those waters.
• 19:10
  Speaker 2: Uh i'm curious to hear about your idea over uh AI regulation when it comes to Europe because that's a big topic.
• 19:17
  Speaker 2: you know though?
• 19:17
  Speaker 2: I talked with the entrepreneurs.
• 19:19
  Speaker 2: is there as the regulation?
• 19:20
  Speaker 2: we're talking umbrella term regulation right for any sector.
• 19:23
  Speaker 2: There's the continuous complaint puts too much regulation on innovation, and we're talking about AI.
• 19:30
  Speaker 2: You've just made the argument that is one of the most important kind of advancements we have had.
• 19:35
  Speaker 2: so how do you see European regulations?
• 19:38
  Speaker 2: We got the AI Act things like where to stand on it?
• 19:42
  Speaker 1: Yeah I think there's certainly an honest to ensure public good is always preserved regardless what technology its power amount.
• 19:53
  Speaker 1: There's also this other conflicting pulling force of allowing the market to do its thing and allow this technology to accelerate.
• 20:01
  Speaker 1: We will be left behind if we don't, it is a difficult balancing act.
• 20:06
  Speaker 1: I think that problem with regulations like the EU AI Act.
• 20:09
  Speaker 1: they are half-assed in there.
• 20:11
  Speaker 1: If you're lawyer worth your money You can kind of circumnavigate.
• 20:15
  Speaker 1: so if going into place regulations into play then need So, definitive and focused enough that they do protect public good.
• 20:25
  Speaker 1: They're there and designed to still allow innovation in market growth with protection of the individual and the state on all the rest not just like a new piece of bureaucracy That you just need to take a bit time sort of getting around To carry on.
• 20:39
  Speaker 1: I think a lot of the conversations i think are most interesting.
• 20:41
  Speaker 1: Are things sort of like ownership itself?
• 20:43
  Speaker 1: Like if someone clones your voice or eleven labs?
• 20:46
  Speaker 1: Do you have a voice as I exact your data?
• 20:48
  Speaker 1: sort of things that people put on the open internet?
• 20:50
  Speaker 1: That's now being used to train models.
• 20:51
  Speaker 1: People have to pay for.
• 20:53
  Speaker 1: but is free data on the Internet?
• 20:55
  Speaker 1: Those are places where i think an individual isolated regulation makes sense.
• 20:59
  Speaker 1: one big overarching regulation.
• 21:01
  Speaker 1: four ai It it's too watery, its too broad.
• 21:05
  Speaker 1: this kind of by saying we're going to regulate science.
• 21:07
  Speaker 1: Okay which bit biology like in euro like?
• 21:10
  Speaker 1: well Which bit?
• 21:11
  Speaker 1: I think these isolated regulations out more scoped and specific have often made more sense.
• 21:16
  Speaker 1: And I think we've seen various examples of that, again using my IoT cloud.
• 21:21
  Speaker 1: all other sort of previous technical advances that happened taking similar trends has happened in the same way.
• 21:26
  Speaker 1: you start with this like overarching idea for regulatory regulations and laws And then usually like shattered into small pieces because we see that this overarching piece doesn't capture all of the nuance of different parts.
• 21:38
  Speaker 1: I'm not a lawyer, so take everything i say with your recommended daily amount of salt.
• 21:43
  Speaker 1: but that's my thinking.
• 21:45
  Speaker 1: always you know as something comes from the world government and public good As it were...I think We shouldn't just have this dystopian welfare.
• 21:52
  Speaker 1: The technocrats can do what they wish for the sake of profit.
• 21:56
  Speaker 1: But..we need to be realistic.
• 21:58
  Speaker 1: And if, for example we see that maybe players and nation states are going to use AI in ways that aren't as wholesome balancing forces, I would say.
• 22:16
  Speaker 1: Which is that's from the most political diplomatic answer you've ever heard to that question?
• 22:21
  Speaker 1: Well
• 22:22
  Speaker 2: i expected nothing less.
• 22:24
  Speaker 2: it was just in line.
• 22:25
  Speaker 2: yeah It's good that these conversations are happening not just between you and me but really with anybody that's innovating in Europe right now.
• 22:33
  Speaker 2: they have it in their mind as how AI can impact what they're doing how regulation is gonna impact what AI's doing to their sector.
• 22:40
  Speaker 2: So if anything, I think it's positive that we're all having these conversations really so early on and then were not kind of waiting.
• 22:48
  Speaker 2: see whats going happen where trying take grab the bull by horns.
• 22:52
  Speaker 2: But with that in mind, I'm curious to hear if because now you've entered the entrepreneurial world and i'm sure You have a more contact.
• 23:01
  Speaker 2: Now And your speaking With more founders such as yourself Have you seen there To be a difference?
• 23:07
  Speaker 2: The way they approach AI and AI safety from When you worked MI six MI five when you're at Whitehall Is There A Difference Kind of Between The Government Approach To AI risk Versus The Entrepreneurial.
• 23:24
  Speaker 1: There was this really old meme that used to go around of like if there is a big internet trend, ten years later someone would put an Internet Explorer has finally caught onto the trend.
• 23:34
  Speaker 1: That's very much government.
• 23:35
  Speaker 1: Government as much I love them are very slow when it comes to tech adoption Obviously, so you kick out when it's this behemoth on an organization that does many things.
• 23:47
  Speaker 1: So also trying to keep on top of what new technical trends are going is a struggle whereas in the entrepreneurial world and startups at the open market It's it's so fast people are constantly on top Of What's New?
• 24:01
  Speaker 1: What's happening especially When it comes to AI.
• 24:04
  Speaker 1: everyone is so engaged and excited.
• 24:06
  Speaker 1: I think The Open Market has utterly eclipsed is happening at the government level, especially in different ecosystems.
• 24:14
  Speaker 1: I think you know famously SF and U S. it's very rapid It's very fast And but i think specially in UK and in the EU there are still this like really rapid speed of growth when it comes to startups on AI and tech.
• 24:28
  Speaker 1: But Government Is slow.
• 24:31
  Speaker 1: The UK has just spun up a sovereign AI fund I Think its sort of five hundred mil size one, which is a good step forward.
• 24:41
  Speaker 1: It'd be interesting to see how the capital has deployed and what their focus will be?
• 24:45
  Speaker 1: Is it going to be infrastructure or are there gonna be application layer sets these sorts of things?
• 24:49
  Speaker 1: but yeah I think that time we spent in SF for New York seeing people would just go crazy with AI was absolutely incredible.
• 25:02
  Speaker 2: Okay, why?
• 25:03
  Speaker 2: and you're building over my now-and-time as we've said where everything says there's more and more companies More and more different focuses on AI infrastructure cyber security You know even if we sector by sectors a I for legal tech ai four aggregate tech, AI for everything.
• 25:21
  Speaker 2: So how do you see Overmind standing among this competition?
• 25:26
  Speaker 2: Now just quick figure.
• 25:27
  Speaker 2: I had a quick look over our coverage and in the past year when we're talking about AI security infrastructure innovation funding We saw around seventy five million euros And that's per our coverage.
• 25:39
  Speaker 2: so How Do You See Overmind Standing In This Area And Distinguishing Itself?
• 25:45
  Speaker 1: Yeah, so I guess one of the key things we never well used to but now never really describe ourselves as a security company.
• 25:51
  Speaker 1: We're definitely in the infrastructure world.
• 25:54
  Speaker 1: It's about being this sort of ubiquitous layer that enables a lot of powerful things on top.
• 25:59
  Speaker 1: So if for example let me say like agriculture health tech whatever it is If these teams decide hey what i want To do Is build technology?
• 26:07
  Speaker 1: That is bespoke.
• 26:08
  Speaker 1: The industry are working In or my very specific use case.
• 26:11
  Speaker 1: maybe For example open AI says you know What am gonna Do.
• 26:14
  Speaker 1: we're going to build an agriculture LLM.
• 26:17
  Speaker 1: Here you go, here's a model focused on LLMs.
• 26:20
  Speaker 1: does it know anything about the fertility or soil in Tuscany?
• 26:23
  Speaker 1: Probably not because this is general models and maybe want to train something of that specific data And thats.
• 26:29
  Speaker 1: where would come in?
• 26:30
  Speaker 1: We allow teams take what is generally intelligence and power it.
• 26:33
  Speaker 1: then Right now people love just say are doing AI for X. It's becoming less and it's interesting to say that you're doing AI for X because there is so ubiquitous.
• 26:43
  Speaker 1: Everyone who's doing AI or something, its kind of like saying your website was on the internet as well.
• 26:49
  Speaker 1: Of course it is!
• 26:49
  Speaker 1: Because...it is now the floor To be using AI.
• 26:54
  Speaker 1: So next step on top of that Is if when a thousand competitors are just these general products That make an API call The newest Claude model How will you differentiate?
• 27:04
  Speaker 1: What would you do that is technically exciting is more powerful than all the other teams standing next to you and a trend that we're seeing at the moment, is proprietary models.
• 27:13
  Speaker 1: Models are trained on data from whatever corporate your working in.
• 27:17
  Speaker 1: And the Intercon team their FIN product.
• 27:19
  Speaker 1: they recently released a fine-tune model Now there like main competitive edge as one of if not only company within this sector That has a proper proprietary model trained on database.
• 27:30
  Speaker 1: shit into The market.
• 27:32
  Speaker 1: so that's what we were enabling.
• 27:33
  Speaker 1: We're enabling teams do these very quickly rather than taking weeks and months, it takes hours for them to train a model so they can then put out into the market.
• 27:41
  Speaker 2: Okay well in this next question is maybe a bit premature you know continuing at seed.
• 27:48
  Speaker 2: but I think that's interesting And i always wanted to ask it to founders who come from another world within the entrepreneur world because You really have found founders.
• 27:58
  Speaker 2: divide those two camps.
• 27:59
  Speaker 2: once I come From kind of real-world application bring it Into uh into new company or one side just go straight into building a new company.
• 28:07
  Speaker 2: Both are success, some were successful and others not.
• 28:09
  Speaker 2: it's up to me decide which.
• 28:11
  Speaker 2: but I do want kind of hear about if there was any difficulty on your end in making that pass over because i think something gets overlooked.
• 28:20
  Speaker 2: sometimes you have to change the mindset or attitude from being an employee to then moving onto lead.
• 28:30
  Speaker 2: my own company.
• 28:31
  Speaker 2: is there any bumps so far?
• 28:35
  Speaker 2: purport.
• 28:36
  Speaker 1: Yeah, I think anyone that does this regardless of where you come from there's going to be huge challenges even if you're going from a really empowered role at like a big tech company stepping into this all of the sudden There are like a thousand things You need to constantly be on top up and i think for me?
• 28:54
  Speaker 1: I guess so i was kind Of lucky in some regards in that i'd come From A very unique environment Where there were just Like a thousand Things each day That you Just had To Be So Cognizant of and especially like the technical sons who are just having to find new ways to solve entirely unique problems every single day.
• 29:12
  Speaker 1: And there weren't really many people that could help, you had a team around but You kind of have to sort work this out on your own which is very much The same as this.
• 29:20
  Speaker 1: or you're coming into...you're doing something That Is gargantuan Like the idea Of trying to found A company if you're Doing VC?
• 29:28
  Speaker 1: The Idea about Company being worth hundreds of millions If not billions.
• 29:32
  Speaker 1: It's Not Small Endeavour.
• 29:33
  Speaker 1: And so stepping into that, you're going to have a whole range of problems.
• 29:37
  Speaker 1: That come your way and for me I think the whenever there has been a bump in the road be it where do we move the product?
• 29:44
  Speaker 1: How can we hire HR visas or all the rest?
• 29:47
  Speaker 1: The best thing is just surround yourself with people who've run this continent before Or could just sort of being on your corner when they need them.
• 29:55
  Speaker 1: We had great advisory team, hired really well.
• 29:58
  Speaker 1: Our investors are brilliant.
• 30:00
  Speaker 1: But there are people now I can fall back to and say, look.
• 30:02
  Speaker 1: I don't know how to answer this question quickly.
• 30:04
  Speaker 1: Can you help me?
• 30:05
  Speaker 1: Answer the question quickly And that is fine.
• 30:07
  Speaker 1: There's absolutely fine.
• 30:08
  Speaker 1: i think The public profiles of founders That you also have told to put forward Is one of sort Of This indomitable person who just knows everything Like theres not a bit of problem I cant solve.
• 30:20
  Speaker 1: It's not true Doesn't matter Who they are.
• 30:22
  Speaker 1: Its absolutely Not True.
• 30:24
  Speaker 1: They Are Surrounded By People That Are Helping Them.
• 30:26
  Speaker 1: at the end of day, you are just one person.
• 30:28
  Speaker 1: So surround yourself with people that can help you
• 30:30
  Speaker 1: win.".
• 30:31
  Speaker 1: That was a generic answer but say yeah they're off course there are a range of problems.
• 30:37
  Speaker 2: Yeah okay well then it's my final question and I'll let you go uh... But its important i think in all these interviews to ask about where do see your company going in the future or what kind of goals you have for, I mean.
• 30:51
  Speaker 2: You've already mentioned that you evolved from AI security into more infrastructure.
• 30:58
  Speaker 2: so do you see another revolution coming down the pipeline?
• 31:01
  Speaker 2: Where where do you want to see over mine five years from now?
• 31:03
  Speaker 1: yeah i think this this sort... The pivot from Security Into.
• 31:07
  Speaker 1: This has been like the most validating thing..I think all of a sudden the company is growing in ways You know, we couldn't really imagine.
• 31:15
  Speaker 1: The open source product is growing very well.
• 31:17
  Speaker 1: the enterprise stuff was going really Well.
• 31:18
  Speaker 1: so now We are sort of locked on to series a first half or next year We're gonna be expanding into us over this here as well.
• 31:26
  Speaker 1: and also the rest of recipe.
• 31:28
  Speaker 1: So really it's how big can you go?
• 31:32
  Speaker 1: Is always in question.
• 31:33
  Speaker 1: And for our sis We'll see, like it as large we can.
• 31:38
  Speaker 1: But yeah the core focus is approaching series A. so making sure that the enterprise adoption is really strong The partners I've been working with at the moment stick and build something compelling for them because we're an open source product.
• 31:50
  Speaker 1: obviously building the community ecosystem you have there's very important.
• 31:54
  Speaker 1: So running a bunch of microphones events all these sorts things over the coming year And be finished hiring now or team has grown nicely So it's heads down and running straight onto series A.
• 32:07
  Speaker 2: Okay, well we hope to hear from you.
• 32:10
  Speaker 2: I'll keep my mailbox open i'm sure.
• 32:13
  Speaker 2: uh...I'd probably even understand more of this for the Series
• 32:16
  Speaker 1: A that is
• 32:17
  Speaker 2: for sure.
• 32:17
  Speaker 2: um okay thank-you very much Tyler.
• 32:20
  Speaker 2: It has been a pleasure to have you on.
• 32:22
  Speaker 2: Thank You!
• 32:23
  Speaker 2: Alright?
• 32:23
  Speaker 2: And hopefully will keep in touch.
• 32:25
  Speaker 1: Yep, far enough so see ya later.