Yevgeny Kaspersky: ‘AI Will Be Pervasive, Like Electricity’

From Ekspert, May 5, 2026. https://expert.ru/intervyu/ii-budet-vezde-kak-elektrichestvo/. Condensed text:

Editors’ Note. – Yevgeny Kaspersky, the founder and CEO of [IT security company] Kaspersky Lab, is known for his passion for travel. In a conversation with Ekspert, he acknowledged that volcanoes are his thing. His Lab is dynamically changing – and is even setting the goal of surpassing the frenetic pace of global transformation. Essentially, Kaspersky remains an active volcano: He has been running the company for almost 30 years now and intends to devote at least another decade to its development. [In this interview] he speaks about the threat of global technological isolation, the transformation of business, and what 19th-century peasants would think of modern urbanites.

* * *

Question. – Should children be banned from using neural networks to solve problems – for example, math problems? Or on the contrary, should they be taught to use neural networks?

Answer. – My children go to a school where students simply have their smartphones taken away: They have to leave them in designated lockers, and the teachers consistently enforce phone rules. I think that is correct. After all, education exists not so much for the sake of knowledge as to teach a person how to learn and to develop a person’s cognitive abilities. Otherwise, it would be like watching sports from the couch and thinking that you’re gaining muscle.

Q. – And what about homework?

A. – At home, we see to it that the kids do homework themselves. Of course, it is strange to assign a supervisor to a child or simply deny access to a phone. Alas, it is difficult to resist the temptation to do puzzles, for example, which were so popular in the past, with the help of a neural network if a smartphone is at hand.

Q. – In other words, your answer is that, for children, neural networks are –

A. – Evil (laughs). Seriously, it’s a useful thing: When you need to learn something quickly, [a neural network] will tell you. But using neural networks to solve math is a bad idea. It’s better to teach your children mathematics.

Q. – In an interview with [Tinkoff Bank founder] Oleg Tinkov (labeled as a foreign agent in Russia) and Oleg Anisimov 15 years ago, you said that thanks to their fundamental technical education, Russian IT specialists were No. 1 in the world. What has changed since then? How would you rank them now?

A. – Ten years ago or more, Condoleezza Rice (US secretary of state in 2005-2009 – Ed.) said to me: “Russian software engineers are the best!” And I would add: “Russian hackers are the worst!” In the sense that they are the most dangerous. I still agree with Condoleezza and think that our IT specialists are better than any foreign specialists overall. And every country has naturally talented people.

Q. – Are you seeing a brain drain from Russia?

A. – There were times when a certain number of people left [the country]. I am currently not seeing a major brain drain from the [IT] sector.

Q. – What will future IT specialists be taught, in light of new technological developments like vibe coding?

A. – I will say it again: mathematics. A mathematician can always grow into something really cool, topnotch: an IT specialist, economist, scientist, businessman – whatever. But this does not work in reverse. As for vibe coding, non-IT people say: “Wow! It writes its own code [and] develops programs!” But let me tell you: There is nothing revolutionary about that. At one time, [they] programmed directly on hardware – literally moving pins from one place to another. Then programming languages appeared – first on punch tapes or punch cards, and only later on the screen. And so on and so forth. After all, the Internet did not come out of nowhere: There were government-funded networks, [such as] ARPANET [Advanced Research Projects Agency Network] and FidoNet. This technology became revolutionary only when it evolved from a niche technology into a global one. However, for those who created it, it was evolutionary. The same goes for both vibe coding and artificial intelligence (AI) as a whole.

Q. – Do you have a virtual private network (VPN), and how often do you use it?

A. – For probably 35 years now, the first thing I do after brushing my teeth [and] sitting down at the computer is turn on a VPN. They are currently talking about this issue a lot, but for the most part very one-sidedly. VPNs were created to securely interconnect remote parts, for instance, of one company – so that employees based in St. Petersburg, Sao Paulo or Hong Kong and so on can work in a unified virtual space with protected data and internal systems. No major company can operate without a VPN service! That is to say, it can work physically. After all, people used to work with the help of snail mail. But then all processes would grind to a halt. I am exaggerating, but in reality, work without VPNs would be far slower.

Q. – Have Internet restrictions affected your business?

A. – A little. The main inconvenience was related to restrictions on popular messenger services. After all, we are a global company: We have 37 offices across the world. But for all intents and purposes, we have our own messenger.

Q. – Have you installed the MAX messenger [on your phone]?

A. – My [personal] phone can only make voice calls. The corporate phone is based on our operating system (KasperskyOS, developed by Kaspersky Lab – Ed.). MAX is still incompatible with it.

Q. – Do you use such a phone for security reasons?

A. – Yes. And practically no one has my number – only the people closest to me.

Q. – All the same, some of your phone’s components were imported. Aren’t you afraid of being hacked? How likely is that?

A. – My phone is built with Chinese components, but its design [and] architecture were developed by Russian engineers. It is impossible to hack it. But this train of thought is correct. Suffice it to recall Operation Triangulation targeting iPhones.

Q. – Could you explain the connection?

A. – To remind you, in 2023, Kaspersky Lab experts discovered the so-called zero-day vulnerability in iPhones. [Attackers] installed graphics chips designed for internal testing and debugging to obtain access to the entire physical memory [at the user level]. In other words, it was not so much a vulnerability as an undocumented feature.

I believe the hackers were 200% sure that even if someone discovered that their iPhone was acting strange, they would be unable to remove the Trojan. After all, the iOS operating system is a black box, unlike Android. We managed [to remove the Trojan]. So when they talk about something as a closed system that is off limits to any external users, that means at best protection against accidental, uninvited guests. But specialists can gain all the access we need. Generally, I think Operation Triangulation was the most sophisticated espionage campaign in the world.

Therefore, to preclude such risks, I prefer to use only domestically assembled security cleared smartphones. . . .

Q. – Other countries – for example, France, Turkey and South Korea – are currently creating or developing national messenger services. Are you seeing a trend here?

A. – I certainly am. Everyone wants data to remain within their national digital environment.

Q. – But you said that software alone is not enough. There will always be hardware loopholes. What is the point, then?

A. – Creating national hardware is the next stage. Only 30 years ago, the IT industry was just an amusing toy. The average person might have said: “Oh, how cool, I don’t need a telephone now – I can call via the Internet!” But now it is critical infrastructure.

Today, the country’s economy [and] security depend on the stability of the Internet [and] cloud data accessibility. By now it’s like electricity. Can you imagine any self-respecting country moving all of its electric power stations to another territory? That is absurd! [They] can simply pull the plug on you at any time.

Q. – It turns out that now any large country will try to shut itself off from others and ensure the full localization of such technologies. What could be the consequences of such fragmentation?

A. – The consequences of isolation are simple. Instead of working together on a common joint project, everyone will go and invest in more or less the same thing. Some will succeed and others will fail. Those who succeed will have to find markets for their sovereign technologies. Otherwise, investments will incur losses. Those who do not find [markets] will see their losses pile up. The result of accumulated losses is basically clear.

Q. – Theoretically speaking, can a given country build a controllable [and] manageable but efficient Internet?

A. – From a logistical standpoint, that is impossible. Of course, if [a country] really wants to completely isolate itself from the outside world, it can switch to snail mail and fax machines. However, in reality, there will always be holes in the iron Internet curtain. Even North Korea, which is considered the most isolated country [in the world], still has global Internet access – at least for a certain segment of the population like civil servants.

From a purely technical perspective, building fully sovereign infrastructure [and] Internet is possible, but very difficult. It requires components, equipment, data centers, cables, optics, software and so on.

Q. – Can Russia at least approach such sovereignty?

A. – In theory, it can. We have the brains and competencies for that, but no market. Our IT projects would not survive without exports; at best, they would eke out a miserable existence. . . .

Q. – In recent interviews, you have often been asked about international business [cooperation]. You have mentioned Latin America, Asia [and] Arab countries among the focus regions. What are your criteria for moving into new markets?

A. – First, we are talking not about regions, but specific countries. Second, we are not moving into new markets – in one way or another, we are already present in almost all markets. There have probably been no new areas for us over the past 20 years or so. We have clients on Easter Island.

So in this case, it would be correct to say that we are refocusing. Our current priorities are Brazil, Egypt, Saudi Arabia, the United Arab Emirates (UAE), Nigeria, India, Indonesia, Thailand and several other countries. . . .

Q. – Do we understand correctly that you intend to switch from antiviruses to cybersecurity?

A. – Overall, cybersecurity includes not only software, but hardware. We are not grabbing at everything. Because, for instance, physical equipment has completely different economics: the cost of materials, components, production, storage, deliveries and so on. For years we have been providing comprehensive business security technologies – from antiviruses to SIEM [Security Information and Event Management] systems. Another example is the NGFW (Next-Generation Firewall – Ed.), which we launched and completely certified last year. As I said earlier, our non-antivirus products are developing much faster. . . .

Q. – Are these products developing through import substitution?

A. – More than that. After all, we never think in terms of only the Russian market, but take a global view. Furthermore, our clients know us very well, so it is much easier to offer them new products.

Q. – The world is actively transitioning to a new technological mode based on AI: unmanned transport systems, dark factories and lights-out manufacturing, and so on. How are you reconfiguring your business to take technological transformations into account?

A. – Our business reconfiguration depends not on where the world is headed, but on the transformation of cybersecurity threats. We try to anticipate what will happen next, and that works. For example, we were prepared in advance for the appearance of mobile-phone viruses or the necessity to protect automated process control systems against hacker attacks.

Another example: As we can see, almost half of hacker attacks start with exploiting cyber vulnerabilities. This coming summer, we will move into the vulnerability management market, which is new to us. We are planning to back up our product with a multiagent GenAI system, which we are currently developing in conjunction with Sber[bank (Savings Bank)]. It will automatically monitor client infrastructure protection levels 24 hours a day.

Or, for example, we are seeing vehicle control interception threats. After all, a modern motor vehicle is essentially a simulator. You do not steer a car mechanically. When you turn the steering wheel, step on the gas or the brake pedal, or pull the handbrake, you send a signal to the onboard computer, which then issues a command to the vehicle’s corresponding systems. This is how cars, trucks, ships [and] railway transport systems work. Most likely, also aircraft, but we have not studied them in detail yet. All onboard computers have some vulnerabilities. Incidentally, I cannot rule out a hacker trace behind the incident where a container ship veered off course, ran aground and blocked the Suez Canal for several days (reference to the Ever Given accident in 2021 – Ed.). In short, [hackers] can break into almost any new car, lock the doors and take you to the ends of the earth.

Scared yet? The good news is that when we saw a threat there, we began to develop protective mechanisms. And ensuring transport security is one of our key vectors today. We are offering an automotive gateway [Kaspersky Automotive Secure Gateway (KASG), a software solution for connected vehicles – Trans.]: It isolates all digital components of a vehicle, acting as a secure firewall for all signals. If someone hacks your car, the intruder will be unable to get to the onboard computer.

Q. – Is it safe to farm out any control to AI if a cybersecurity system like yours is installed on an object?

A. – My answer is yes and no. After all, apart from the possibility of direct hacking, there are AI malfunction risks. Plus, there is always the threat of AI being [deliberately] fed bad data and flipping its lid.

Q. – Do you personally use neural networks?

A. – No. When I need to draw a picture, I know who to go to.

Q. – How many people do you have working on AI?

A. – We have dozens of people engaged in AI development projects. But to reiterate: Specialists in this field should not be confused with some virtuoso vibe coder. These are not “AI experts,” but specialists with a solid mathematical background, since AI is primarily about algorithms [and] mathematical models. Some even have advanced degrees – mostly in technology, physics and mathematics. These specialties will always be in high demand.

Q. – So far, investments in AI frontier models are generating losses. OpenAI projects a $14 billion loss in 2026. Are you seeing a bubble in this market? When could it burst and what would happen after that?

A. – Yes, the market is overheated, but this often happens. Remember the dot-com bubble? The same could happen with AI: Bang! And the market crashes. Or maybe it would be like with Bitcoin, when the situation develops in a wavelike pattern. But whatever the case, [this] technology will not go away.

Q. – How do you see the world with AI in the next 20 or 30 years?

A. – I can say that even if a terrible catastrophe happened through the fault of AI, this technology would not be abandoned. In other words, there will not be a repeat of what happened, for example, with the passenger airship industry (the 1937 Hindenburg disaster in the US effectively killed the industry – Ed.). AI will be pervasive, like electricity. Although I would change the term in Russian: This is definitely not iskusstvenny intellekt [artificial intellect]. The [English] word “intelligence” has many meanings. For example, razvedka [espionage and reconnaissance]. One very important area of our business activity is called precisely that – Threat Intelligence, [or] kiberrazvedka [cyberintelligence]: We collect and analyze information about threats and share these expert data with the market. However, in Russian, the word intellekt has one meaning, and it does not correspond to the capabilities of modern AI systems. . . .

Q. – When AI takes on most of the work, will humankind sit around doing nothing?

A. – From the 19th-century peasant’s perspective, a modern designer does not do much heavy lifting: He doesn’t plow, sow or reap. From the standpoint of today, maybe people will indeed [look like they] sit around doing nothing, but in reality they will simply be doing something else. There is nothing new here. We had the First Industrial Revolution, the Second, and now it is the Fourth. Life will become easier, [and] there will be more time for hobbies. And mathematicians and programmers will be in increasingly high demand.

The industrial picture 20 to 30 years from now will be approximately as follows: Let’s suppose a plant is running low on materials. The machine sees this and understands that there is a need for cement, steel, nails and so on. Next, an algorithm places a bid on an electronic exchange, chooses the closest and most optimal supplier in terms of price-quality ratio, and at the time specified, the enterprise receives the required materials, which are delivered by drones. All of that [happens] automatically. . . .

Q. – Finally, a philosophical question. The main principle of AI’s work is the principle of determinism – i.e., complete predictability, cause and effect. However, a person can act irrationally – for example, save another person at the risk of his own life. Can a machine behave ethically?

A. – Saving another person at the risk of one’s own life is in fact rational.

Q. – Why?

A. – Irrationality is not about choice contrary to circumstances, but about random choice. So your question is not so much about ethics or morality as about spirituality – i.e., a kind of inspiration from above. There can be no spirituality in modern AI models. Because no one knows for sure how the human brain works. According to some scientists, it is a set of electrical signals, but they are not predetermined. This is why we live in a nondeterministic world governed by chance. Of course, if we combine five [different] AI models so that they compute differently and install a random switch, that could be a step toward [understanding] how the animal brain works. But even in that case, [AI] would still be far away from a human behavior model.

Q. – You have been running the company for almost 30 years now. Do you plan to leave the Lab and become, for instance, an ordinary shareholder?

A. – I will definitely work here for another decade or so. . . .