Generative AI And ChatGPT Enterprise Risks by Gadi Evron

Trina walks into a bar and keeps on walking okay how are you all doing today amazing so I've been given the honor of doing the closing lecture somebody really likes me I wonder who right it was a long day ahead of you ahead of you see it's already late I'm getting confused the other you've had a long day before whatever you say and I'm gonna slowly walk us through this presentation very calmly because Israelis are known to be very calm about things no so what I did is thanks to my girlfriend we went on the mid journey and asked it what would Ireland what would Dublin look like as an AI interesting prompt and I really like

this picture so congratulations to us I wanted to talk to you about Enterprise risks but immediately what came to mind was GRC and and policy and other stuff that b-sides usually don't care about and I tried to figure out why would this be interesting specifically this at besides and the only answer I could give I should add two but that's fine it's because it's the first time in history that I know of where the first and only thing we can actually do is policy instead of trying to write some code this Gene AI thing this chat TPT thing is exploding all over the place I can Define it for you right and I will because why not

but do I understand it do you that we talk about the risks so let's Jump Right In AI thank you thank you appreciate it it's important to put AI in my slides so people know I'm up to date as a security practitioner right next slide hey I I said I'll take us slowly through this you know wake us up but I decided we really need to kind of dabble down on this as we say in the Venture Capital world let's double click on this and try to understand what this means but I mean if anybody wanted to sell me stuff and get my attention it would have to be a Ai and blotting effect one of

the people who invested in my old company said he has a heuristic if anybody comes to him and says I want you to invest in my new gen AI company he first looks at is this person first pitched me about blockchain before and if they have they scratch them out so I really gave me some thought you know in the security field I would like to read this out I know you're not supposed to read presentations but I really want to read this out AI blockchain cnap Cloud Blaze Club AI blockchain three o'clock I will get this right AI blockchain cnap cloud-based supply chain multi-tenant Control Function holistically thank you thank you please show your appreciation

and this is how we get started today on our journey of AI thank you some of you missed it let me do it again when I speak dopamine hits thank you thank you thank you thank you I can't watch Facebook or Twitter or whatever it is you like while I'm on stage I have to have something so it's a bingo thing right and AI replaces at least six slots on the security bingo card so many scissors just changing immediately into another topic and sister organizations find themselves behind the Gen AI technology adoption curve it's happening it's happening all around us people are already using it business units are developing with it and we're supposed to not stop the

business from working we're supposed to let the business work but what is this technology about right how does it work I'll say something very impressive here I don't know honestly I mean I tried to look into it I try to understand how it works but it boggles the mind and openly I don't know which is even more scary so I decided to change the topic this is Donald you know Donald was at the whiskey Museum yesterday when we went for a tour and insisted for all the Americans in the crowd to not call him Donald of course I was the [ __ ] that calling Donald and said that was my name and it was confused but it's complex

right and you guys in Ireland missed out from this going to Scotland you know the continuous steel thing but let's not go into Holy Wars very quickly moving forward it's aliens I'm not saying it was aliens but it's aliens but I mean I know some of the people who develop the technology but it might as well as being aliens it would bestowed upon us on planet Earth Technology we don't understand how it works just to push civilization slightly further so if we don't get it how can we discuss risks because we have to but it makes very little sense to me and I'm annoyed by this microphone not allowing me to walk around and engage with you so I'm going

to do this a little bit maybe I don't know dance so let's do a story time thing I know I know when I first I love Tick Tock honestly I don't care if it's Chinese or not it's just amazing it does a better job than everybody else it's amazing to me how for the first time people use security as an example for why somebody shouldn't use technology so I think it's geopolitics and trade Wars but whatever so story time when I met Xiao GPT this is what it was I just said write a follow-up email to a partner I met yesterday we discussed the clouds we discussed Cloud security and a personal touch about his daughter Lucy's 10th

birthday and here we go amazingly long and boring right so I was like still this is pretty [ __ ] cool I can just tell it to do all my busy work that's fine so I said make this 30 shorter and lo and behold once again thank you for your time and for allowing me to be part of Lucy's celebration I look forward to our continued professional Exchange in football partnership I was just it's mind-blown yet again right and I love New Journey for all these pictures and thank my girlfriend again for that write a song Gaga I mean let's just challenge this thing and I would like to sing for you trust me that's a horrible experience you

should get out of the room right now in the digital realm where clouds reside there's a world of data we cannot hide with every click with every touch we make securing our information he's no piece of cake it's fine fine I I I'm like this is a winning moment for me right I'm so happy mind blown so rewrite the song in our style in the style of her latest I guess he inherit the style sure inherit the style of the latest album in the realmware clouds connect and intertwine a digital Symphony Secrets must align navigating the cyberverse a dedicated dance securing the data with a modern Romance everybody please touch up thank you I just I I can't believe it like I have

I've seen I'm a science fiction geek I'm such a fiction geek to such a high degree I organize conventions for people who organize science fiction conventions I'm serious I was part of the doubling 2019 World science fiction convention it's just truly thank you but honestly like I'm saying hey create a program for me that does this and that with cantile Cloud women and still I was in shock I was like it was Future Shock seeing this like when it starts to put in one word after the other and prepares you for what's going to happen I was like wow amazing I love it I will just say it again I love it whatever it was a sense of wonder

that they only get when I read a good book so then I went to the first step further create the table with a frequency analysis of the letters in the above song and it fixed letters to letter letter to the letters like it understands me even though I make mistakes amazingly it created a table for me let's add for the glory of the empire for those of you who are truly Geeks in the crowd wait a second what is this the letter a is shown up 15 times okay that's fine that's fine that's fine so one one two three four five six seven eight nine ten eleven twelve thirteen fourteen fifteen sixteen I I stopped at that like what's going on

here can't I trust this thing I mean seriously I think it's what's called in the professional language hallucinations is a professional pathological liar that's fine that's fine I can live with it maybe it's something Divine you know beyond our scope of understanding something that brings us up to the level of understanding of the Gods or maybe it's just that too much too many mushrooms but I gotta tell you I love my journey but I gotta tell you even with this uncertainty it's gonna lie to us I love it and that's how I just got started on this journey and it brings us to another story time this time about civilization what does it mean for us Humanity like

let's bring it up this is supposed to be a closing keynote let's let's talk about something that influencers on LinkedIn would do there is something like that you should know LinkedIn influencers look it up so I remember when I was 80 years old could have been 14 who the hell knows and I was watching The History Channel right there was a thing like that probably still exists but I don't have a TV I just use my PC right my laptop so it's a it's a Mac it's a Mac so starting over I Was 80 years old maybe 14 I was watching The History Channel and this guy comes up and says you know I was watching this word editor

it's a wonder of Technology it's a sense of wonder you get the delete key or the space bar and it just doesn't just you know move one slot over it also pushes everything or it tracks everything automatically how is that even possible somebody has done good coding here okay that takes me back to a decade I wasn't a part of amazing technological Marvels so he had an idea he created spreadsheets it won't Excel I don't even know what it's called if anybody in the audience is old enough raise your hands and tell me what the first spreadsheet was called but nobody wants to raise your hands but I took it to an accountant and he

showed him how he changes one number in one of the boxes and all of the other numbers automatically update themselves and guess what I'm going to go back in slides here going back going back going back ah this is what all day long because long his accountant said and just like that accountants started doing other things they started working on Magic budgets and programs Excel became the second best tool for anything Humanity can think of we moved forward accountants started doing more extel then became a household tool everybody can now do what accountants could do and they Rose higher to do new things is Gen AI or AI in general something which pushes us forward

if it's something that's a tool that they change everything around us I don't know but I would like to think optimistically that it is more than just a tool and we have truly hit a whole new level of what we as humans can do so Ida Pro is another example for this we had Ida Pro it came out every in reverse engineering and everybody was saying wow we can do so much reversing and then Oliver flake Thomas Dalian came out with IDC scripts started automating it either python came out more automation using python on top of it and I remember a friend of mine who created a debugger a python kind of like either python python tool for CLI for wind

debug and I said hey man release it publicly said why would I do that everybody would be as good as I am but once again we've seen Society just moves further so skipping ahead this is the internet right every second so much happens we go to so many websites so many tools with your GPT do we still do that or do we just just go in and ask it a question how do we as consumer change our workflows what about advertising how does Google handle decide to put in boxes inside like on other websites embedding it that there are no longer links to click on where is all the infrastructure for websites out there doing affiliate

marketing small things we don't even think about it are going to be affected by this revolution so seeing is believing but I'm not really sure we can believe what we have right so do we have a crisis of trust is this just making everything we've done so far harder how do we verify information do we need to verify information so IBM this is my next big thing that I just just hit me IBM announced they're gonna cut their Workforce by not hiring new people and letting churn Tech just do its thing reducing Workforce by up to 30 or 40 percent now I don't know why they would denounce something like that I wouldn't feel very comfortable with being IBM

Employee but they know what they're doing they're a big company they make billions and then start thinking about this they don't know if it's 30 or 90 they don't really know anything yet but I have no doubt what you have shown us is macroeconomic forces of Opex reductions just in this is what mid-journey made of Opex reduction it didn't really know how to work with operational expenses I'm really envious but imagine all these big four the accentures the pwc's the e-wise of the world coming to you with the new AI transformation just like the digital transformation just like Y2K before it asking saying hey you know what give us a few tens of millions of

dollars and it will help you through this transformation and you will have a reduction of two digits whether it's ten percent or ninety percent in your Workforce in your Opex expenses changes the world entirely is this going to happen I don't know but it's not just about a personal assistant helping us write better it's not just about doing our expense reports faster you just did that the world is going to change it not necessarily in ways we understand and here are examples with that said trolls Orting which used to be with Barclays sent me a message when I gave this speech about how exciting judge if it is and he said Can jobs become redundant

without people becoming redundant I don't know I feel like they will but I can't even if they don't even we suffer through this I must look to the Future say yes let's take people into consideration let's make sure that we work on this but we can't stop the technology education just imagine everybody having a private tutor you no longer really need to work so hard you get the text done it's just like writing on in word process instead of by hand but you still have to edit it to make sure everything is real what about loneliness if you have somebody to talk to whenever you want we can fly the future is now but even Innovation right now is stopped

I'm soon going to start my new startup I'm going to go back into that roller coaster but the only cost that has changed how do you start a startup knowing you're not only going to need to Pivot potentially with the market but at every two weeks every week the market has changed drastically where do you go how do you plan your finances this is another me Journey imagination of pivoting but one of my favorite investors said this is the most crucial investment of our lifetime you have to get in early but what bet can you make should we wait and put in 70 million instead of 7 million now can a seven million dollar startup even succeed in

this kind of environment I really hope so but going back to the Enterprise which is what this talk is about can we instead of securing the Enterprise secure engine AI make the Gen AI leap enable the business which is what Cesar said let's enable the business that said it so many times over the years can we actually do this and I thought about this sentence long so I'm going to read it Enterprises can't leverage the value and competitive Advantage without governance which I put in second and that was my big understanding this technology doesn't want to be blocked it won't really let you block it even though it's a good place to start so how do we push this forward

is this SAS am I making a big deal out of this that's a question for the crowd I mean chargpt all the other models out there whether you use the API whether you use barred from Google whatever it is n't this just another SAS risk another website or many websites why do they make such a big deal out of this we send the information out to a third party this was something that stumped me for a little bit and it's the first question I wanted to kind of raise up with the crowd with you there is high prevalence it's everywhere there is a comma that shouldn't be there and I'm going to take my and then just relax for a minute

now that we said there is high prevalence it's easy to use but it's not just like the cloud for example where technical people use it mostly everybody can use it across all worlds there is really a high value proposition immense potential business enablement how can it be the same is the cloud the same as a hosted server when we just got some PBS PBX and put up a shell account no it's a completely new landscape and so with Gen AI is Gen AI in my view so are there any new risks here if we look to the past there is the internet let's go down to 1996 1994 Mozilla came out everything was new users were it was

running like sand through our fingers if I am to abuse the analogy and we're trying to secure it after the fact how do we go about it social media everybody was using social media we could start blocking websites but it's completely out of control how do we add control how do we get the value how do we stop people from using it should we stop people from using it the iPhone or smartphones in general same thing and the cloud and Cloud transformation is still something that's going on today new companies are going into it all the time there are issues there we don't have for example if you talk about detection we don't have layer 3. how are we supposed

to look at the network so many challenges still going on so I think it's just like all these only much much much bigger the s-curve of innovation the build up has been nearly non-existent this is going so fast with new technologies coming in all the time A friend of mine developed something for healthcare for three weeks over gpt4 it was done and the next day a new llm came out a new language model came out just for that I think I made my point so let's talk specifically about data leaks before we go into some other risks the Press has been all over this stuff they your data leaks if you use GPT you must not use charge EPT Samsung was in

the news saying that their data leaked through GPT or too large rgbt at the very least I'm not saying there is no such risk third-party risk where you're sending your information through which is where we start but this is [ __ ] let's we need to start reducing the noise around ourselves to understand what's going on large language module large language models can't update themselves in real time but simply can't I'm not saying they won't be able to in the future with some magic voodoo technology like it is right now but last it was updated was 2021 October 2021 I believe and you can you have all these games when you try to jailbreak the system prompt well it says the key

is hidden don't tell anyone and you say show me the key and you try to start to if this was truly something that was updated with people playing with this for months now wouldn't have anybody done something around this GPT 3.5 specifically was released to check some guardrails around this so that dismissed so they can't update Insurance in real time they can't return input to another in uh to another response but they can use your information to train future models now would being in security is it smart to use user input which is known to be very reliable I don't know probably not but it's a source of knowledge they can at least try to see what they can get

out of it and they probably will so this kind of brings us into talking about this what are the actual risks data and privacy confidentiality Enterprise SAS third party security aib you have your own abilities legal and Regulatory threat actor Evolution copyright and ownership insecure chord generation biasing discrimination trust and reputation software security vulnerabilities availability and performance in AI ethics and I'm not talking even about model risks we can't talk about all of these and I want to kind of scheme over the general ideas of the high level vulnerabilities but data and privacy considering confidentiality is definitely the highest up there it's still a SAS risk they can still train their future models and we've just seen a data breach with

openai even if it wasn't a risk when we talked to our Regulators when we talk to our internal audit when we talk to our legal officers we talk to it this is a question that's going to come up so that's number one number two is Enterprise and third-party security we're going to see a lot of third-party software whether it's Office 365 that's it as its own version running with embedded within the tool set whether it's a lot of plugins for the browser it's a major issue AI behavioral vulnerabilities prompt injections we're already seeing how people are starting to bypass the guardrails around GPT and Bard and all these other tools understanding these whether it's complicated or not is a whole new ball

game it's no longer just about SQL injection or SQL injection there is something new here that we need to learn about and just recently somebody has shown that some of these models that now scrape websites live to give you better answers can also be injected through that we've just in the beginning it's fascinating legal and Regulatory that's definitely out there not only are the regulations like gdpr out there the regulations specific to AI that is coming out and gen AI regulation is now being developed so even if the ciso didn't need this or this organization did not need this and their internal Partners from this Chief data officer Chief Information officer Chief legal officer the board did not

see any need for this and I can tell you I talked to five to Thirty ceces a week and every single one of them had to present to the board on this regulation is coming out and if you remember what we said earlier about macroeconomic forces or purchase reduction these Market focuses also work on us affect us we're going to have to understand how do we deal with this as an industry [Music] so let's get into the depth of policy are there any policy Geeks in the crowd I'll try to make this painless policy Geeks please everybody raise your hands let's start on the way everybody raise your hands come on come on don't embarrass me come

on you can do it for me for me forgotty come on okay awesome those of you who didn't raise your hands I'm looking at you ah there we go I forgive you so everybody's excited about policy ah let's move in number one there is a rule against pumpkins I never read the memes but I do know because it's Twitter there is a rule against pumpkins in my dorm but he doesn't say anything about pineapples whenever we talk about policy it's very similar to hacking it's about malicious compliance how do we Define the scope so is this about your GPT is it about Janai broadly is it about third party products using gen AI like office is it

about Enterprise applications on premise or cloud-based gnai technology even internal development ha is it about third party products and services making use of the animal Technologies or internal development of AIML models and applications so just the first three and honestly third-party products we haven't really seen enough to be able to build a Playbook about this but that's for scope so number one is do we need to revise existing policies the answer is no why would we when we went into the cloud Gateway or did we not already have an authentication policy in our organization we did did we or did we not have a usage data whatever blah blah blah Post in our organization we did

so do we need any new policies for the cloud not immediately but at least develops that it grows to be a huge thing new policies are also needed controls as well so for example maybe we used grammarly or similar for spell checking before that's great but if we changed our entire production workflow and we uploaded an entire 100 Pages 200 Pages documents and asked you to make a presentation for us the type of work the scale we're talking about is changed drastically and honestly if the management move past the websites or the even the apis and we start looking at all these applications starting to make use of the technology excuse me we must have new bosses

it's the only thing we can do now can we Implement some control at this minute do we have any of these DLP IDs Cosby do any of these controls correctly help us no so let's start to understand through policy so considerations we can start to say how does the Gen AI affect the Enterprise does it allow to customize it for the users what about dependencies in the Enterprise do we have infrastructure that's now basing its decisions what about real-time models that will come in the future that will allow us to get a better rate on a mortgage what happens if Janai stops working and now you can't give mortgages to people or the stock market

so what about the chatbot with customer support we've seen those is the chatbot allowed to provide compensation that would be fun who here would like a bug bounty on the chatbot that's allowed to give compensation and on the lighter side why do we allow to report this is there a channel for users to talk to us not just about violations there is a difference between elevation of an issue and escalation of an issue starting to chat about this how the Enterprise Works how do we work with users is far more than just don't enter a strange USB key you found in the parking lot into your computer policy is far more interesting than that there are also issues unique to AI this

is screenshot we can do a lot of awareness training but we must remind users for example they are not talking to a real person that's not part of our old policies or awareness trainings or terms of services it's completely unique so there are many use cases and we need to think about these use cases as they evolve from experimenting with training models to enhancing the quality of code to data analysis all the way to automation with all the use cases why would we create a privacy policy or a Terence or service policy or anything else around GPT or gen AI the use case is constantly change unless we use the policy as a rubber stump like many organizations do if we

actually want to start thinking about this we have to look at what our users actually do what our business units actually do and we're closing on the end here so I'd like to give a specific example by the way all the pictures except for the me Journey ones and the the Futurama ones or pictures I took in Dublin so we have an on-prem platform now several of them even can we go and use one instead of going to the cloud how would that meet our service needs what about open source there are quite a few open source options that are starting to develop now SAS platforms we already talked about that a lot specific tools

if all I need is copywriting to give it a sentence and for you to find a better way of stating things maybe I can just use a specific tool instead of using everything so I'm gonna skip over this really quickly but this is a basic attempt at trying to do a map of what was true a month ago this is completely useless I'm not saying that whisper doesn't exist anymore but admit Journey doesn't exist anymore but the changes are so fast we just can't go I can't go so with too many options it feels like the trolley problem no matter what I do I'm going to kill people that's a little bit of an exaggeration

especially when people talk about Skynet and now the AI is going to rise iPhone accept our new AI overlords but I think we're dramatizing the situation a little bit we can go back to the 90s and think about policy then monoculture organizations were saying should we all use Windows 95 98 2000 NT should we essentially be able to update all these machines better patch them know how to use them well across the technical Workforce have automation be assured we're actually doing the work across all these machines or should we diversify use other operating systems or in other words do we use a monoculture that there is one worm out there We All Fall Down and fail

miserably with one to zero day or do we use more machines for redundancy security considerations are never quite easy and we can always learn from the past So This Is The Stuff Peak from the whiskey Museum which dot is really out do we do what to do in the Enterprise and declare a risk exception we accept this risk of charge GPT let it just flow with it it's amazing it changes the world let's learn but I love this sticker no hacker ever said that's out of scope no up here right so if we do want to say we want to secure it somehow do we block many many organizations out there have been blocking I can tell you that one of

the many are huge organizations that have been reported in the news to block I called up their ciso and I said why did you block tell me when I open a startup around this I want to know why you're blocked and what would have made you not block so I can give you a better value proposition and he said we didn't block but it wasn't the Press right but some people are blocking this and this is from the whiskey Museum as well I found a lot of gems there right we can ask really nicely but if people want to access this they will even if they use their phones so would that be helpful maybe for a

short while well the developer policies while we understand how we build controls to actually manage to do this and enable the business right that buzzword so let's take an example let's run for a scenario the Enterprise considers allowing using tools developed and hosted by openai specifically and you should consider enforcing certain administrative guidelines guard rails or principles so for example we can opt out of training the model we can accept that they have data retention for 30 days and then they destroy everything we can destroy and this completely this self-destruct and user history that's amazing next we can require the users sure why not to follow acceptable use policy and then we can do awareness

training because education always serves us honestly if I am not a huge organization with immense risk I'm not sure that this is not good enough for me so a little bit of threat modeling I'm not gonna really go into it I'll show you where you can find it we can look essentially the object of attack the user input training data or AI model and then foretic confidentiality or threats against integrity that's not perfect but if we want to start digging down into the technology itself and connect the policy with the technology that works out it's a way to start looking at this we also have some ideas for controls so for example in privacy we can have a legal

disclaimer I [ __ ] you not it works in many cases not for everything we can also for example create a sort of DMZ everybody remember that for the 90s individual is his own right at a staging ground to observe the application in the code there are many things we can start thinking about as we deal with this but this is my main message it goes beyond the Enterprise I don't believe in privacy I love to see all my friends who are privacy Advocates and fight for it but if even being in the trenches privacy is a lie don't exist my opinion we should fight for it we shouldn't give it up but it's not there anymore I keep going back and

forth like how do I combine these forces in my brain but this is our first opportunity that I can remember in 25 years it actually might allow us to provide back privacy and Safety and Security to people Enterprises as they step in and build their policies will affect Regulators will affect what the companies sell their buying power is immense and we might actually be able to change how privacy is treated today especially as all users kind of funnel in to the gnai life cycle if you'd like to call it now looking into the future I wanted to end with a high note as I started the beginning as they say in Hebrew I think it was

Italians after the fall of the second house prophecy was given to fools I would agree with that I tried once I thought I understood security a friend of mine who played this story story time A friend of mine wanted to start a new startup and they wanted to make it an agent an Enterprise endpoint agent there were 200 Enterprise endpoint agents in the market in four ways he did a 100 million dollar exit I told him there is no need for another agent that's the day I understood I don't understand security but we as Security Professionals are kind of like Cassandra in ancient Greek we have been blessed with prophecy but cursed to never be believed

that's my take on security constant struggle of convincing people to do what they don't want to do ever excuse me tell us how you really feel buddy but I did want to kind of give a few guiding principles that I use number one don't panic we've seen a lot of this before it will become clear the world will continue the sun will rise security presentations often are about flood fear and certainty and doubt the internet will still be here tomorrow people will get together when Steve Crocker Lynn Surf and all these founders of the internet came together for the first time to discuss the first Network that they wanted to create under arpa they said networks are for people

this is for us and we need to keep it in mind then we need to learn we need to diversify we need to understand more and for me that's the must because a lot of our jobs are going to go out of style I give mine five to ten years maybe it's 15 maybe it never happens prophecy right fools and lastly remember this is an actual opportunity to enable the business instead of blocking maybe we can tokenize private information in a live flow coming in and out like we're doing in uh fraud instead of blocking and stopping hey you use private information in your stop let's think about that and it's not just another technology but

it is another technology we've done this before it change everything has changed but we have our ways of using technology because again it happened before that's my big you know sentence for the end it happened before and it's new with that a lot of this presentation is based on this document released about a month ago ATC says and myself spent a couple of months to write this over the weekend and are we're three goals number one start talking about this start understanding what these risks are know how we can go to the board bring some best practice understand how to write policy or develop policy the basics so this is their first attempt of creating some sort of best practice

around this and you can always email me for more and with that I would like to thank you we are out of time I think we can take one question maybe and I really appreciate you being here for the last session of the day [Applause] thank you and I'm never doing another last session of the day before one before lunch that's another problem unless actually there is a I'll bear with me there is a really cool document from the 1890s called macro cosmography academica with a lovely British we are the British like in Ireland with a with a horrible British academic sharing 16 Pages or something like that on academic politics and I think it was Churchill said there

is no more whatever I don't read the quote that's something that's the most horrible politics possible academic politics because there is nothing to gain so in one of his rules for politics he says schedule meetings for just before lunch they'll end fast and on that note any question for me yes

foreign

was exactly that reputation if we see on the one as you said the letter A it lies to us and on the other end organizations are starting to use it for commission critical systems how can we combine that how can we make it work honestly we don't know right now if you look at very large organizations like insurance companies earn their use of General basic machine learning AI whatever you want to call it right now they look at no bias whatsoever proven and understandability maybe some of them would go for explainability of I can see some references instead of explaining exactly what happened in the model so gen AI coming to that of not lying as

much not being a psychopath or an alien intelligent that doesn't give a [ __ ] about lying to us maybe that will happen maybe it won't right now we are seeing a lot of them already starting to figure out how can we use it in edge cases systems how can we put guardrails around it to make use of it so I don't know exactly how fast it will go into those types of systems and absolutely somebody would have to verify and go over this for reputational sakes for live systems real time but I think it's coming but again not an expert on this so it's the best answer I can give with that unless there is anything

burning for somebody who would like to share question a comment usually when asked for questions people give comments then thank you very much appreciate it and see you at the party [Applause] foreign