Invisible machines: nanotech, AI and the end of anonymity

In the 1990s, nanotechnology was sold as a coming revolution — tiny machines building anything atom by atom. That revolution never arrived. Something quieter did. Nanotech became ordinary industry, hidden inside almost every device you own and a surprising number of medicines.

Chips: the biggest nanotech application of all

Today's transistors are measured in a handful of nanometres. TSMC started mass production of its 2-nanometre chips in Taiwan at the end of 2025 — its first process using gate-all-around transistors, where the gate wraps the channel on every side to stop current leaking. The 1.6-nanometre A16 process is due in the second half of 2026, and A14 in 2028. Every AI accelerator, data center and future 6G modem rests on this progress.

Coatings: a Finnish invention

Atomic layer deposition — ALD — grows coatings one atomic layer at a time, and it's now essential to chipmaking. It was invented in Finland by Tuomo Suntola in the 1970s, who received the Millennium Technology Prize for it in 2018. Finnish companies Beneq and Picosun became well-known equipment makers; Picosun is now part of the American giant Applied Materials.

Medicine

The best-known example is the lipid nanoparticle — a tiny bubble of fat that carries mRNA vaccines and some gene therapies into cells. Nanoparticle-packaged cancer drugs have been in use since the 1990s, and many rapid tests use gold nanoparticles to produce a visible colour.

Everyday products

Zinc oxide and titanium dioxide particles in sunscreen, battery electrode materials, dirt-repellent surfaces and the quantum dots in QLED screens are all nanotechnology. Nobody calls it that any more.

• Nanorobots in the bloodstream. Labs have built folded DNA structures — DNA origami — that open only when they meet a specific cancer cell, and magnetically steered microrobots. Promising, but far from routine use.
• Wonder materials. Graphene and carbon nanotubes were supposed to transform everything in the 2010s. They've found specialist uses, but cheap large-scale production has proved much harder than expected.
• Self-replicating nanomachines. Not a real research direction, and nowhere on the horizon.

The risks are known, not mysterious

A substance can behave differently as a nanoparticle than as a larger lump — it may reach the lungs or pass through tissue. That's why the EU's chemicals regulation, REACH, now requires separate information on the nano-forms of substances. It's a recognized, regulated risk.

And one myth deserves a straight answer: there are no tracking chips or network-connected nanobots in vaccines. The nanoparticles in vaccines are fat. They contain no metal, no antenna and no electronics, and the energy and communication problems of a truly networked nano-device remain unsolved even in the best laboratories.

Here's the key point most people miss: most nanostructures have nothing to control. A nanoparticle has no computer, no antenna and no battery. Its "program" is its chemistry and its shape. So control happens in three places — in manufacturing, in design, and through external fields. AI is now central to all three.

1. Manufacturing — AI is already here

When a transistor is a few nanometres wide, the smallest deviation ruins production. Chip fabs are among the most data-driven factories on Earth. AI pre-distorts the masks used to print chip patterns — light's wavelength is larger than the features being printed, so the patterns must be computed — and machine vision inspects wafers for defects no human could see. Thousands of sensors track temperature, gases and layer thickness, and models adjust the equipment before a deviation can form. That is industrial IoT in its purest form: sensors, data and AI in one closed control loop, and it's everyday reality.

2. Design — AI as the inventor

AI now helps design nanostructures before anyone makes them. Models predict how new compounds will behave in batteries or catalysts, so only the most promising few out of tens of thousands reach the lab. Protein structure prediction — the work behind the 2024 Nobel Prize in Chemistry — makes it possible to design new molecules. And AI is used to tune the recipe of lipid nanoparticles so they reach the right tissue. This is where AI will probably accelerate nanotech the most.

3. External control — the remote

When something small has to be steered inside the body, it's done with fields, not a network connection: magnets guide magnetic particles, ultrasound or light triggers a drug's release at a specific spot, and DNA structures can be designed to open only when they meet a particular molecule — the logic is written into the structure itself.

What about the "Internet of Nano-Things"?

It's a real research concept: nano-scale sensors talking to each other over terahertz frequencies or with molecules, the way cells signal with hormones. It even overlaps with 6G's sub-terahertz research. But it is still mostly theory, because nobody has solved how a nano-sized device gets energy, sends a signal, or works in tissue that absorbs terahertz radiation. In reality, IoT and nanotech meet the other way round: nanomaterials make ordinary sensors smaller, more sensitive and more power-efficient. Those sensors are still normal-sized devices with normal radios.

In 1991, Terminator 2 gave a generation its image of the future: a liquid-metal machine that could melt, flow and reform. Remarkably, this is one of the few places where that fiction has a real research counterpart.

The figure that melted out of a cage

In January 2023, researchers from the Chinese University of Hong Kong and Carnegie Mellon University published a "magnetoactive solid-liquid phase transitional machine" in the journal Matter. It's gallium mixed with magnetic particles. Gallium melts at 29.8 degrees Celsius — almost in your hand. An alternating magnetic field heats it into liquid by induction, magnets pull the liquid where it needs to go, and when the heating stops it hardens again into something solid and load-bearing.

In the experiments it jumped over moats, climbed walls, split in two to get around an obstacle and merged again. The famous demo was a little Lego-style figure that melted and flowed out between the bars of a jail cell. The detail that usually goes unmentioned: it didn't reassemble itself. The camera was paused and the researchers recast the liquid in a mould. This is not a self-reforming T-1000.

The applications the team named include removing foreign objects from the body, delivering drugs and soldering circuits in hard-to-reach places. The nearest to practical use is electronics repair — flowing into a tight spot and solidifying as conductor and solder at once. Medical use is still far off; the senior author himself noted that biocompatibility needs a lot more testing.

A whole field of liquid metals

• Stretchable electronics. Liquid metal inside rubber channels stays conductive while stretched or bent — ideal for wearables and soft robots.
• Shape-shifting antennas. Change the shape, change the frequency — interesting for 6G, which needs tunable antennas.
• Self-propelled droplets. Researchers in China showed a liquid-metal droplet that moved on its own, powered by a chemical reaction, back in 2015.
• Cooling. Liquid metal is already used commercially as a thermal interface between processor and cooler, for example in the PlayStation 5.

Why the T-1000 is still fiction

Two things separate the film from the lab. The first is intelligence in the material: every fragment of the T-1000 knows what it's doing. Liquid metal contains no computation at all — switch off the external field and it's just metal. The second is its own energy source: moving, heating and changing shape take a lot of power, and in the demos it comes from laboratory electromagnets. The closest real field is "programmable matter" — small robotic modules that can join into different shapes — but those are still millimetres to centimetres in size and need power and communication.

The next few years: invisible gains

Chips shrink a few more generations, but the focus shifts to stacking and packaging, because shrinking alone is no longer enough. Liquid metals arrive first in specialist roles — cooling, flexible sensors, wearable electronics and tunable antennas. AI-driven materials discovery speeds up new battery, catalyst and coating materials; possibly the most important change of the period, and the least visible. And the success of mRNA pushes lipid nanoparticles into other treatments, such as cancer vaccines and gene therapies.

The 2030s: soft robots and precision medicine

Soft robots with liquid-metal "nerves" and "muscles" may spread into rehabilitation, prosthetics and robots that work alongside people. Medical microrobots may reach their first clinical uses in tightly limited tasks — still doctor-steered tools, not autonomous machines. Self-healing materials may appear in electronics and structures. And quantum and nanotechnology converge, because quantum computers and sensors are themselves nanotech products.

Beyond 2040: open questions

Programmable matter is a real research target, most realistically as systems of small modules rather than a single liquid. Building things precisely molecule by molecule is advancing slowly, mostly through biology — designed proteins and DNA structures. And an intelligent, autonomous liquid-metal machine is not on the horizon at all; physics offers no current path to putting computation, energy and communication into every particle of a material.

In Minority Report, police release small robotic spiders into an apartment building. They crawl under doors, find every resident, open their eyes and scan their irises. How far away is that? Broken into parts, surprisingly close — just in a different shape.

• Small autonomous robots: real, but flying. Insect-scale robots exist in research, such as Harvard's RoboBee, but they carry little and run briefly. In practice, drones took the spiders' place. Pocket drones weighing around 30 grams have been in military use for more than a decade, and indoor drones that navigate on their own are commercial technology.
• Working as a swarm: partly real. Coordinated exploration and mapping by several robots is active research and partly in military use. Closed doors and stairwells remain hard.
• Iris scanning: real at close range. It's used at airports and in national ID systems, but usually needs short distance, good lighting and cooperation. A robot prying eyes open is still fiction.
• Recognition at a distance: real, and far more widespread than in the film. A camera and face-recognition software do the job from across a street, with no robot required. Gait recognition works from even further away — and even Wi-Fi signals have been used to tell people apart by the way they walk.

So the film's purpose — finding and identifying a specific person in a building or a city — is already achievable. Only the method differs: cameras, drones and networks instead of spiders.

In Europe, the limit is law, not technology

Since February 2025, the EU's AI Act has banned police from using real-time remote biometric identification in public spaces, with tightly limited exceptions such as searching for a missing child or preventing an imminent terrorist attack, and only with authorization. Remarkably, the same law also bans the film's other big idea: predicting an individual's risk of committing a crime based solely on profiling. "Precrime" is explicitly illegal in Europe. Elsewhere it's different — face recognition is widespread in public spaces in China, and in the United States practice varies from city to city.

The film's subway scene — cameras reading every passenger's eyes as they walk, billboards greeting them by name — may be its most accurate prediction. Much of it is already real.

Face recognition in the metro: already here

In October 2021, the Moscow Metro launched Face Pay, letting passengers pay by looking into a camera at the turnstile, and rolled it out across the whole network — the city called it the first large-scale system of its kind. It drew heavy privacy criticism, and Moscow's metro cameras have reportedly been used to identify protesters. Airports, meanwhile, already run corridors where travellers are identified biometrically without stopping.

Iris scanning on the move: works, under controlled conditions

Here the film is still ahead, but not by much. Research on reading irises from walking people goes back about fifteen years, and commercial systems work at a few metres. But people have to pass through a defined gate or corridor, the scene needs near-infrared lighting, and the distance is metres, not tens of metres. It doesn't really matter: face recognition already does the same job well enough. The iris is more precise; the face is visible from further away.

Billboards that know your name

Screens with cameras already exist in malls and stations, but they typically estimate a viewer's age or gender or whether they're looking — not who they are. Naming someone would require a face database, and in Europe the GDPR forbids using biometric data for identification without explicit consent. In practice, the same personalization already happens another way: your phone's location and browsing history tell advertisers far more than your face. It just happens on a screen in your hand instead of a wall.

The eye transplant: the film's deepest insight

To escape detection, the hero has his eyes replaced. That's the fundamental problem with biometrics, dramatized: you can change a password, but not your face or your iris. If a biometric database leaks, the damage is permanent. Real-world evasion uses infrared glasses that blind cameras, patterned clothing that confuses algorithms, or simply masks — and the algorithms keep learning to get past each trick.

Put all of this together and one pattern repeats. Nanotechnology itself is not the biggest privacy threat. What it enables, combined with AI and cheap sensors, is.

The chain

• Nanotech makes sensors and cameras smaller, cheaper and lower-power. When a sensor costs a cent and runs for years on a battery, it can go everywhere.
• AI removed the old bottleneck. Surveillance used to be limited because nobody had time to watch thousands of cameras. Now every image, sound and sensor reading can be analysed automatically and combined with other data. Collection was already cheap; understanding is now cheap too.
• Wearable and on-body sensors — flexible electronics on skin, in clothes or inside the body — collect data from the body itself: heart rate, movement, sleep, stress. Health data is the most sensitive data there is.
• Drones and small robots make surveillance mobile. You can avoid a fixed camera; a moving one is harder.
• Biometrics is where it all meets — and it's the one kind of data you can never change.

From the device to the body

Until now, tracking has targeted devices — phone, computer, card. You could protect yourself: leave the phone at home, harden it, pay in cash. The new tracking targets the body itself: your face and iris on camera, your gait on camera and in radio signals, your body's reflections in a sensing 6G network, your physiology in wearable sensors. You can't leave your body at home. That's why many of the old personal defences lose their power, and the weight shifts inevitably from individual choices to law.

The next frontier: the mind

Taken to its end, the same trajectory leads to brain–computer interfaces, whose data may reveal emotions, attention and perhaps one day intentions. Chile added so-called neurorights to its constitution in 2021, and data measured from the brain has started to receive its own protection elsewhere too. It's likely to be the biggest privacy question of the 2030s.

What protects us

Law, above all — and Europe's is the strictest in the world. The GDPR treats biometric and health data as special categories whose processing is prohibited by default. The AI Act bans real-time remote biometric identification by police in public spaces with narrow exceptions, emotion recognition in workplaces and schools, and predicting crime risk through profiling. But the EU's proposed Digital Omnibus reform, still under negotiation, will decide whether that protection holds where AI is concerned.

Technical design — processing on the device, collecting less, sensing that outputs statistics rather than individuals — is possible, but optional and usually more expensive. Individual choices are limited but not meaningless: choosing devices and services deliberately, using your GDPR rights, and taking part in the conversation while the rules are still being written.

Science fiction got the direction right and the shape wrong. There are no shape-shifting killer robots and no spiders in the stairwell. Instead there are materials that are quietly getting smarter — conducting, healing, sensing and adapting — designed by AI and produced in factories run by AI. Less dramatic than Terminator, and probably more far-reaching.

• Nanotech won by becoming invisible. Its biggest impact is the chips that make AI, data centers and the next networks possible.
• AI controls the very small in the factory and on the design table — not through a network inside your body.
• The T-1000 had a real moment in 2023, but liquid metal is a remote-controlled tool, not a thinking machine.
• Minority Report is largely here — just with cameras and networks instead of spiders.
• The privacy question has moved from our devices to our bodies, and next to our minds. No phone setting solves that. Standards and law will.

As with 6G, the technology itself doesn't decide whether it protects or watches. Standards, laws and markets do. The difference now is what's at stake: not just the data on our devices, but our faces, our bodies and eventually our thoughts. Which is exactly why this is the moment to draw the lines — while they can still be drawn.

Henri Isola

The Code Breaker · Holistic & Business Coach

Twenty-five years as a Solution Architect and Senior Technical Expert in enterprise data centers and cloud migrations — IBM, HP, Sanoma, CGI. Now reading human systems instead. Also read 6G and beyond: what's really coming.