J3 FA 02 Hugh Karp Insurance for smart contracts

hi everyone my name is Hugh Karp and the founder of Nexus mutual so a little bit of background on myself I'm the I'm the corporate insurance guy that caught the blockchain bug a while ago and I've now kind of worked out how to do insurance on on blockchain you know in a mutual fashion so I'm an actuary I've got 15 years experience in the insurance industry just left my nice cruzi job doing CFO for Munich Gries Life operations in the UK for those of you don't know Munich Reiser one of those huge companies that no one really knows about so it's a reinsurer it's got 33 billion of premium income so I guess that kind of tells you two things hope hopefully it gives you some confidence I know what I'm talking about when it comes to insurance but secondly the market for insurance is absolutely huge and so and blockchain is a good way and it's always been talked about as one of the major use cases of blockchain so hopefully this is this is one way we can we can make it work so I'll start with a I start with a bit of a story so 2000 3000 BC Chinese trading merchants used to basically put their goods on boats and go down the river and basically enabled trade that's how they moved their Goods around and but they had they had a problem as a as a community they had a problem they those often rapids and they'd and basically the goods would get lost or destroyed etc so so as a community they basically grouped together and and the merchants decided to shift different goods on to different boats so they would basically all share their their goods around on the different boats so they're if one went down there they only lost 1/10 or whatever of their goods and so that meant that meant quite a lot of things it was it meant that as an individual merchant you were more resilient you could you're only exposed you're less exposed to being totally wiped out you could also instead of only purchasing some Goods and then transporting them you and keeping a reserve yourself you could take more risks you could use more of your kind of capital built-up funds and and take more risk and then enable trade a bit more so so this is a this is like one of the first examples of mutual insurance of the community recognizing a problem coming together sharing that risk together and becoming more resilient and then enabling greater trade and everyone being better off so that's kind of what we're what we're trying to recreate now that we have watch-chain technology so that that model was good but one of the big problems was it couldn't really scale very much so we put in place insurance companies to essentially link different groups linked the Chinese trade merchants with the I don't know other village over here that was insuring life insurance for different people so those two groups couldn't talk to each other before but now we have blockchain technology that allows them to talk essentially together without the need for an insurance company to link them and become that intermediary so so that's what we're trying to do so I'll start with the product and then I'll go back to the platform and kind of pattern how and how how we're operating it so a Syrian community that's the community where we're targeting first and the problem the way trying to solve is smart contract security so it's obviously it's a real barrier to adoption we've got if you're if you're a user how do you know that if you're trying to interact with the smart contract how do you get that confidence that the money's just not going to disappear if you're a developer you're gonna be can be quite scared of had actually hitting that deploy button to main net and going a hold on you know if this has a bug in it we've gone through security audits but if this has a bug unit what are we going to do so it's also a really hard problem and there's a lot of really smart people working on this but you can go through testing you can go through security audits multiple layers you can go through bug bounties but you can never be 100% sure that there's not something there so from my perspective we need redundancy in the in the solutions as a community so and this is where the combination n solutions comes in play and essentially what I'm saying here is we've got the prevention which should be the primary focus which is what a whole bunch of really really smart people in this room and know in this conference are working on and we're not doing that but formal verification testing tools standardizing code security audits all of that it all needs to happen and we need to get better at doing that and that's great what we also recognize is that you can never get a hundred percent satisfaction or comfort that all of that will work even with as an example even with formal verification if you can prove that the specifications operate as intended and you actually go through that formal verification process that's great but there's the other part that's missing and that's the validation if at the end of the day a when we're coding stuff we have to get our intentions out of our heads and write it on pieces of paper or in code or whatever and then you have to write those specifications so you can test at the specification to get validated and operators intention as work as from you intended to but what's what about that first step did it come out of your head right and did you write specifications correctly so we're saying that prevention should be what we're absolutely 100 percent focused on but we also think we need a cure a safety net if things do go wrong okay that's that's great but okay how do you actually assess claims at the end of the day what we're doing here is we're introducing a human layer on to the code so what we're covering against is unintended usage of the smart contracts that leads to a financial loss and that's really a human interpretation so as an example like the Dow hack or the parody multi-sig well at issues I think it was pretty clear in all in all those circumstances that that wasn't the intention of the code it wasn't the intention of the Dow to be able to it was the intention to be a like autonomous investment vehicle in a whole bunch of different projects it wasn't the intention that someone should be added just drain it funds same with the parity multi-sig issues obviously the intention to keep on secure wasn't the intention that someone could just hack it and take all the money so you can argue whether code is law and all the rest of it and that's fine but I think it's pretty clear that the intention of the code didn't help work in that case and that that's the idea that's what we're trying to cover against and you can't code that you have to have people to assess it so we're providing that human layer and so we're providing a decentralized claims assessment method to actually assess that now that has a whole lot of benefits it has some disadvantages as well one of the big advantages is that its product agnostic we can actually write our product on a piece of paper so the product that we're building smart contract cover can just be did it have unintended code usage that led to financial loss and people can interpret that how they will so it doesn't require a trust of an external Oracle we're not restricted to the parametric covers and that's that's quite that's quite important because parametric will become bigger and bigger and bigger as an insurance aspect of the insurance industry over time as more and more data sources become available but at the moment most of the covered written the vast majority is actually indemnity based which means if you've got your house so there was in part at you the kitchen burned down you didn't get paid for the whole house you get paid for the you lost part of it not all of it and then that requires interpretation by humans to be able to do that one the other quite important thing is because we are voting and everyone's a member of this mutual we're legally not classified as insurance in in the UK and in other countries which means we don't actually need an insurance license or the regulatory framework around it we are essentially just sharing risk amongst each other so this is an insurance it's called discretionary cover and it's provided on a mutual basis these these entities exist today what actually we are doing is taking it a finally legal structure automating it as far as possible with smart contracts and and then just fitting in with that existing legal structure so disadvantages it's harder to implement you've got a voting system with incentives and game theory and all the rest of it it's potentially more vulnerable to attacks and the one thing that we haven't been able to solve just yet but we're hopeful in the future but is that we're not fully decentralized the while the voting mechanic and the first part about is the claim valid is decentralized with the voting the problem is that insurance claims assessment is quite hard and it's very you don't necessarily want to punish people just for having a difference of opinion but you want to punish people heavily if they vote fraudulently so determining what the difference is between are you just got a different opinion or is fraud that's very hard so we're centralizing the decision-making on was that a fraudulent vote or not we think there are solutions in the future but we're going to start a bit simpler where the centralized bid to start so I'll talk a bit more about the platform and then kind of how it works in each of the different aspects so we're trying to replace that insurance entity with code so at the moment we've got a current model of the insurer where you've got all of these services that are grouped together and they provide all of those so you've got your own rating your pricing your capital what you do without capital investment management product development governments governance claims it all all that's all them under 110 and so what we're trying to do here is break it apart commoditize each function and make it very cheap to operate so we will automate what we can and then essentially be centralize the rest and we'll use token incentives to essentially incentivize the work of for example people in this room to provide work to the systems make the whole thing operate so we will automate the sum of the under-18 the investment management will be automated the claims assessment as I discussed that'll be decentralized the capital provision that's essentially decentralize effectively a crowd fund and then we'll have voting processes on governance and we'll also have a the pricing and underwriting is actually split some of its in the middle some of it's like an automated algorithm and some of its decentralized and I'll get back into it into that in a bit so so the big one of the things here is that we will need a whole bunch of external work to be provided to the system to make it operate so we have to incentivize those through through token token incentives so one of the clear ones here is capital provision insurers need a whole bunch of capital to operate they essentially sit on that money and pay it back to the members or the to the people that are insured in the future the claims so this is a kind of a well-known one in crypto obviously you can have a token model and crowdsource of money so one of the interesting things that we're trying to do will actually use a continuous token model so we'll be constantly open and so the price will keep going up and basically it will go up slowly until we until we've got enough capital to back the contracts that we need and then it'll start going up very steeply basically saying we don't want any more money don't give it to us we've got enough to fund what we want and then over time if we're up when we write policies we go I hold on a minute we've got lots of policies in force therefore we need more capital therefore we'll drop the price of the tokens to encourage more capital to come in so it's very dynamic and the model is trying to have an efficient capital allocation and making sure that we only raise money when we need it and we don't over ace so if you compare that to say a just a regular like ICO where you go well we'll raise like okay 50 million and then we may not be able to go back at a later point what happens is you over raise capital to start with because you might need it in the future but then you can't go back at the end necessarily so what we're saying is insurance companies need ongoing capital as they grow and this is a much more efficient way of deploying that capital in addition will you'll be able to lock your tokens to participate the surplus distribution so for basically we generate profits and we don't need any more capital then we hand out back to the members that's this year basically this whole thing's a membership company instead of the profits going to the shareholders they go to the members of the effectively Dow claims assessment so as I said incentivize voting basically you stake some tokens as bond and then and you can lose them if you vote fraudulently but it's kind of like the mining operation and a mutual basically you vote and then you get more tokens if you vote with the consensus and writing so well the obvious questions here is how the hell do you price this thing how the hell do we determine what price we should charge for contract smart contract Davis my contract B and all the rest of it so we'll have a two-tier pricing approach represented by those those two lines so the first tier is essentially a automated algorithm that's very conservative that's trying to replicate how battle-tested is this code how long has it been on Maynard how long what funds does it have how long has it been exposed to a hack and it hasn't been hacked and so basically the price starts high very high probably uninsurable so not very helpful to start with but then over time it becomes more reasonable so that's obviously going to work for some contracts out there that have been on there for a while but it's not going to work for a whole bunch of other ones that's where the second layer comes in and here effectively what we're doing is incentivizing crowdsourcing smart contract security expertise so if as an underwriter that's my contract order to say you have the skills and you want to assess that particular smart contract and you think it's secure they new can stake some value against it in the form of tokens and for every sale every cover that's bought on that particular contract you essentially own commission income if there's a bug in the if there's an early bug then you stand to lose some of your stake but otherwise you and essentially commission and as a result the price drops to a much more reasonable level so that that's kind of how we're trying to crowdsource the under-18 expertise and this actually mimics they approached like Lloyd's of London use in their in their models they basically have capital providers that will back an individual expert who under acts a particular table big to play insurance so for example if you're a smart contract or little bit you don't have much capital then someone may back you and that and then you can come to an arrangement with them - about how you split that Commission so we're trying to mimic through smart contracts how the insurance industry works at the moment in addition there's a governance function basically this whole thing is a it's a down it's completely upgradable and changeable etcetera so you can you can vote on proposals you can put proposals forward and that would be like product upgrades changes to parameters or in an extreme event if things go really bad and you can vote to shut the whole thing down and basically distribute the funds - to the members and the membership does have full control over this thing one of the interesting things is what we're doing is there's a both of their legal structures that actually allow this to happen and so you can write articles of association for a regular company and have it operate in this way and this is how a mutual is kind of operate at the moment so that's kind of just quickly kind of buildup how it goes through to kind of give you a little bit more idea of how it fits together because there are basically a lot of different elements that come together to make the whole thing work so there's one capital pool where the way the money goes in their backs the contracts and the first step is as a member you're put in ether to the pool and there is a token model and you get some tokens so that's like basically you get more tokens at the start and less tokens at the end when it's well capitalized then okay you've got your tokens what can you do with them so the first one and most obvious point is you actually destroy them to purchase some cover on a particular contract so you just beat you're just a regular customer and so if you just want to be a regular customer you don't really care whether it's a token price model is because all you're doing is we were buying ether then you're for tokens and then you're destroying tokens they gave cover so if you take those first step to the end all you've done is put in ether and got cover so you don't care what the price is and so everything here and this kind of goes back to what Christophe was saying earlier and the volatility of crypto currencies volatility is actually a relative concept so at the moment because spike and what this is one of the reasons we're starting with this product is we're going to denominator cover in ether because that's generally what people are interacting with smart contracts so we're going to have a pool of ether backing of assets backing its liabilities in ether so volatility is not an issue one ethers 181 ether so because because of that we're just totally kind of within the cryptocurrency space rather than having to worry about the Fiat aspects we we've built the platform to actually be capable of taking in like stable coins once once they get a bit more established and therefore being able to do right however that would you need in a real world for your washing machine is an example so but we'll start with something that's denominated in ether anyway so okay so you've got your cover that's that's one option that's what you can do we've got the pricing model that sits mostly off chain so second one you as I explained before you lock your tokens for participating the claims assessment and if you vote with the consensus you get more tokens and then the claims come out of the pool and go to the member who had had suffered from the hack be more complicated now so you can lock key tokens for surplus distribution now the idea here is as with the token model the we're going to have this off chain capital model so the way insurance companies work at the moat at the moment is they have to work out how much capital do I need to hold to be very confident that I'm not going to go bust and there's a whole bunch of regulations to work out how you do that and there are standard models that for example solvency - there are standard models that tell you how to do that how to do those calculations what we're doing is replicating that exact approach and we have to run that off chain because it's just way too complicated to put one chain so that introduces trust elements you know what I guess from our perspective we want to be fully decentralized but we can't quite get there yet big as a technical limitation so we'll start with something that has some centralized elements and then try to improve it over time sorry going back to the capital models you've got this capital model basically says this is how much you need to hold based on the pulses you have and then that determines your minimum capital level and then we'll just do a ratio say 180 percent of that minimum capital once the funds in the pool get to greater than that then we'll start distributing surplus kind of like a dividend but legally it's a surplus to the members so that goes to the members based on the token Holdings essentially yeah you can lock your tokens to get access to that and the idea here with one of the key themes you'll practice e running through the the token model is that you need to insurance how is all about risk sharing and risk sharing happens over time risk is incurred over time so if you lock tokens then you're exposed to that risk over time so the whole idea is if the more you lock your tokens the more rewards you get out of the pool in return for taking on more risk yep and then the last one is being able to lock them for under aiding so actually there should be a feedback loop into the pricing model effectively but they say yeah you lock your tokens as I was describing before as the stake on the bond and then you earn Commission income in the form of extra tokens so basically if you want to just be a customer here and just purchase tokens burn tokens and get some cover that's that's fine you will be a small member in Yorba still be entitled to some of the future profits but you will effectively get slightly diluted by everyone else who's actually a more active member of the member of the mutual providing work to the system so every one that's providing work to the systems get gets rewarded with extra tokens which is essentially a share of the of the combined pool so it's his balancing act between that you can play on the passive side you can be more active you can choose what you want to do you can choose to be any of those options and so if we take that back to what the the insurer insurer model is basically we do everything in one box going back to that slide from a little while ago what we're doing is we're breaking it out and allowing different experts to participate as they see fit in the items that they're more skilled at doing so we can monetize those functions and basically strip out massive layers of cost from from the whole system at a high level the the economics of insurance works that basically you only get sixty five seventy percent of your premium backing claims the rest of it gets taken away through extra costs in the system and profits to shareholders etc so based on this model we conservatively we think we can strip out at least half of that tax and the rest of it actually just goes to the members because no there's no hold it here that's all everyone owns it together so that's basically it we're using the smoke contracts to to bring back the peer-to-peer mutual we we will when we launch which is hopefully be before the end of the year but will see the we will will launch with the live product and and then the essentially it will the first stage will be the the raising of the capital pool so it will launch it something live and you'll be able to use it from from day one yeah so that's us thank you is there is any question and hello thanks for the presentation you mentioned your products your product includes governance mechanism and that people will vote on a grade on which contracts use and everything can you talk a bit more about that how are you going to structure this is it only going to be vote will you have administrators how will you you have a model for the governess you're going to be using yeah we will use a very simple model to start with basically there will be a legally we'll actually have a board we're a real company that sits around this so there'll be a board of say five people that effectively will whitelist the proposal so anyone could propose something but they get whitelisted by the the board and then the first step is that the board will assess the proposal and provide a recommendation basically a yes or no what they think and then and then they will push that to the membership base to decide what they want to do so every proposal will come from a recommendation from so called experts and the but then the membership base can decide what they which way they want to go in it and it's a mr. token weighted vote the I guess there is there are a few extra elements to it but one thing is that the board will be able to and this is another slightly centralized aspect the board will be able to do some things by itself primarily emergency actions one of the things we have implemented is a an emergency pause button which basically just says stop all transactions happening to basically if there's an issue it'll give us time to build something and then switch from this smile contract to this one over here that's actually fixed if there's a bug so there will be some kind of emergency actions that the board can implement by themselves without pushing it to a full member vote so all right thanks everyone [Applause]