Hi everybody. Thank you. It's a pleasure to be here with all of you today. I want to talk a bit about privacy in the age of AI because we know about AI but privacy is also a very important topic and sometimes these things are at odds. So I think today I want to talk a bit about Proton, about what we do, who we are, so you get a bit of background. Let's talk about AI and privacy, what we've been doing in that area, and also a bit about what Europe can do in order to succeed in the AI age. I think that's also a very important topic because we are, as we've seen yesterday maybe earlier today, not really in the lead. So a bit more about Proton. I'm a scientist actually, I'm a physicist who came from Taiwan to work at CERN, so that was what brought me here. I was academic for the first part of my career, and Proton was really born at CERN in 2014. Our vision is very simple: we want to build an internet that actually puts people first, that serves the interests of people. This is something we started back in 2014. Back in 2014, it was a very different era. In those days, our experiment of building core internet services in a privacy-first way was sort of a crazy idea. Proton today is sort of known for not having venture capital funding. It's not really by choice. Back in those days, nobody wanted to fund this because this was kind of an insane idea. The idea was that you want to compete against Google, the world's biggest tech company, by offering services that they give away for free and get people to pay. People didn't really believe that was a possible business model. But we kept working, we kept pushing, we kept building out service by service. Today, our ecosystem actually contains quite a few different services. Our goal, 10 years ago but even today, is really to become the trusted European alternative to Google and to really help Europe have its own tech stack. I think this is very important. What makes this different as well is that all of our services are open source and we actually utilize encryption to ensure privacy and security, not by marketing or by design but actually by mathematics. This is a bit of background of what we do. This model over the past 10 years did become relatively successful. It started very slow, but as awareness of privacy issues increased in the past 10 years, we also got more and more users. Today, just in terms of numbers, Proton has 100 million users, but also what is very interesting is over 100,000 businesses that also use our services for better security, better privacy. We employ 500 people, almost all of them here in Europe, and today the company is still headquartered in Geneva, so actually not very far from here. Okay, so that's a bit about us, what we do.
Let's maybe think about the main topic here, AI and privacy. And maybe we do an experiment to kind of wake everybody up. So if you use AI, please raise your hand. Okay. And for all people who have your hands up, how many of you trust Sam Altman with your data? Yeah. Okay. And for the ones that do trust Sam Altman, how many of you trust the person that Sam Altman has to listen to, which is Trump, actually, right? because he runs the US. Okay, I don't see very many hands. So this demonstration in a nutshell is basically showing what is probably the biggest business opportunity for Europe in the 21st century because AI is something that everybody is going to use. But fundamentally, most people do not really trust the businesses today that are at the forefront of AI. And this is a massive opportunity for Europe because we are the continent of trust, security, privacy. I think this is the opportunity. So I want to show a bit what Proton has tried to do to take advantage of this opportunity and a different approach of doing AI that we think is more private and more secure, but it's still in its infancy. There's a lot of work to do here in this area. So let's take a look. Last year, if you were following the news, we actually launched something called Lumo AI. The core concept is how can we push out AI into the world in a more privacy-friendly way. The technology to do it actually is very difficult. It's not entirely there, but there are already a few things that can be done. Lumo has five or six things that make it different. Number one, you can make AI more private by simply not recording and logging things that people are doing. So not logging information without permission. This is already a huge step. We can also not use users' private data and private conversations to train the model. This is also better because it reduces the risk that something you've said in AI gets dumped out as an answer that AI gives to somebody else. You can also have a ghost mode where actually nothing in the conversation is ever saved or retained. You can also encrypt the communication between user queries and the GPU. So actually if you can encrypt everything, something that you type into Lumo, Proton servers don't have access to it, and no one can read it until it gets to GPU. This is also good for security. Then the most sensitive thing for AI, AI is a bit like search in some sense, a lot of people use it for search. We all know that your browser history, your search history, is something that you don't want to keep around. A lot of people say when I die I want my browser history automatically deleted. Well, it's similar for AI. What you can also do to make AI more private is you can use zero-access encryption. With Lumo, all the conversations that you put into it are encrypted in a way that we cannot decrypt. So that's also another guarantee of privacy. I think also very important is you can just base your AI in Europe where there's better privacy and security laws and where you don't have things like the NSA and the FBI that can go and get your data and do what they want with it. This is a bit the approach that we've taken. Of course, there's a lot more that can be done. There's also new encryption that will come in the future that will allow us to even encrypt the prompts when they get to the GPU. This is the next frontier of what we call confidential computing, and in the long run can actually get better and better. I think what Lumo shows and the fact that it's successful and people use it is that there is a demand for privacy-first AI, and companies that exploit this demand and work on this actually tap into a very big business opportunity that most of the world has not really realized exists.
And then I want to transition to another thought, which is why do we do this? I think it's also very important that we create European alternatives. We talk a lot about sovereignty, we talk a lot about news about this, people saying we shouldn't buy American services, we should buy European tech services. But actually, I think it's important to talk about why. There are probably two or three reasons. Number one, today I don't think there's actual sovereignty without tech sovereignty. The classic example that we can give is Greenland. People were very happy when Trump said I'm not going to use force to take over Greenland, but actually he can take over very easily. All he has to say is tomorrow in Denmark, Apple, Microsoft, Google, and Amazon will stop working. And if it says that, they give up Greenland. So today there's no difference between tech sovereignty and actual sovereignty. That's the first point. The second thing is tech sovereignty is not just geopolitics, it's also economic. Why? Because if we don't develop AI, we don't develop a tech industry here. It's a bit like in the 19th century missing the steam engine. You simply will not become economically irrelevant and your continent becomes impoverished over time. So if we want to have an economic engine of Europe that works in the future, building a tech sector, building AI, this is absolutely essential and something we need to invest in. It's an investment in our future. The classic example that Europe has made in terms of tech is we've always seen technology as a cost. And if you view it as a cost, the answer is to outsource it to the lowest bidder. That is typically going to be the US and Chinese companies that have the first-mover advantage, which have managed to achieve scale faster. But technology is not really a cost. Technology is actually an investment in the future. And if we want to invest in the future, it's very important that we actually build technologies here. The key part of that is not actually supply, it's demand. How do we create demand for European services? To give you a statistic, today in the average year, the continent as a whole, Europe, we are spending around 200 billion procuring US tech services. So this is our money, our wealth that is indirectly being used to subsidize American tech companies who come back here and compete in a relatively unfair way. Also, a lot of that money that is being spent is actually government money, taxpayer funds. So the reason there is a very big movement for buy Europe these days is because the logic is we probably should not be using our wealth, at least 200 billion per year, to subsidize foreign economies who do not come here and pay taxes and even do not create that many jobs because data centers don't create many jobs. So what does this require? I think what happened in Switzerland is we need to have competent policymaking and competent politicians. Many of you may have heard the rumor that Proton may be leaving Switzerland because of regulatory issues. I want to give one example of what poor decision-making looks like from a policy standpoint and why it's very important that we do the right thing. There's a proposal right now to introduce NSA-style mass surveillance in Switzerland. This of course is really bad because it undermines our claims for security and privacy and Switzerland's reputation. But business is ultimately competition. What is very important is look at the competitive implications. The proposal, of course, they want to have mandatory data retention. So even if you're not guilty of a crime, they want to retain your data. They want to track IP addresses. They want to enforce government backdoor where the government can get access to any Swiss tech company's data. That of course is a huge security problem because it means that the security of companies like Proton and other Swiss tech companies is at the same level as the government. If they can hack the government, they can get into our systems. And e-government is pretty easy to hack. And of course there are human rights issues because this information, a lot of it including IPs, will be shared without a court order. This is the current Swiss situation. The proposal for companies with more than 1 million users is actually to enforce all these obligations. And if you think 1 million users is not a big issue, in fact the proposal will also hit companies with 5,000 users. To give you context, Proton got 10,000 users on the first day after we launched in 2014, so we would have been in scope on day one. Now, what does Europe do? This is our neighbors next door. Let's go to Germany, just across the border. Germany, in fact, has nothing. And then you think, well, how about the US? There's the NSA, there's a lot of surveillance and spying there. Actually, the US also has nothing. So my headline is wrong. It's not NSA-style, it's worse than NSA-style mass surveillance. The most interesting element to look at is what is the obligation that is going to be imposed on US tech companies operating in Switzerland, which are much bigger than all of our local companies. Actually, also nothing. So the summary is, usually when countries make regulations, we make regulations to protect our own companies and keep out foreign competition. But we're the only country in the world that's dumb enough to create regulation that makes our own companies less competitive within our home market. This is one example of regulations that come out of the political sphere which don't really make sense, that ultimately come from a lack of understanding of technology. We also saw this recently with some recent proposals about AI which, had they gone further, would have destroyed the Swiss AI industry.
So this is just one of many examples of regulations that come out of the political sphere which don't really make sense, that ultimately come from a lack of understanding of technology. And we also saw this recently with some recent proposals about AI which, had they gone further, would have destroyed the Swiss AI industry. So what is the answer? I think the answer is, as researchers, as people working in the field, we need to do more to communicate with politics. We need to do more to be at events like this. We need to get the word out there because if we don't have a voice, these types of accidents will happen and they're happening with more and more regularity. So let me conclude by giving you one thought. I come from Taiwan originally, and I think we also need to ask in the long term what is our edge? What is Europe's edge? I can tell you as someone who is from Taiwan and knows the culture there, we're not going to work harder than the Chinese. It's not possible. We can try, and trust me, I tried here as well within Proton to get us to work like the Chinese or Taiwanese do. It's not going to happen. We also are not going to have more money or a bigger market than the Americans. It's not here. We are too fragmented and we're too small. So in fact, the only advantage that Europe has in competing in tech in the 21st century is trust. Trust is the one asset that we have, and trust is also the same as privacy. So if we want to succeed in the 21st century, let's double down on our advantage that makes us competitive and let's defend this advantage. So thank you very much.
