AI concerns and what to do about them
Artificial intelligence makes scams and fake content easier to produce. Fear alone protects no one; a routine does. For each major concern you will find what you can do right now, and where the limits of what we offer lie.
Where the figures come from: Czech Association of Artificial Intelligence, Czech AI Concerns Barometer 2026 (Český barometr obav z AI 2026, a survey in the Czech Republic). Email questionnaire, 1 July – 31 August 2026, 1,407 adults. The barometer does not claim a representative sample, so we write “respondents”. We checked the percentages against the PDF report on 16 October 2026. The barometer on the ČAUI website (in Czech) · Full report (PDF, in Czech)
Try the AI message checkKriticky naživo: talks and workshops
What works today and what doesn’t yet
Working today: courses (most of them free), the Manipulation atlas, the Link X-ray, the AI message check and Send a message.
Not yet: Kriticky naživo talks start with a pilot in December 2026, in Czech in Czechia. The AI check only reads text, not voice, photos or video.
Plenty of worry, little confidence
72.3%
of respondents are worried about deepfakes and fake content
26.4%
believe they would recognise a convincing deepfake
13.2%
believe so among people who do not use AI
This is a self-assessment, not a test. So we do not promise to teach you to spot every fake at a glance. Nobody can do that reliably, not even “AI detection” tools. We teach a routine that works even when a fake cannot be recognised.
What a cloned-voice scam looks like
The voice can sound exactly like a grandson. What gives it away is what it asks for: money now, and keep it secret.
Call from an unknown number, the voice sounds just like a grandson
“Hi Gran, it’s me (note 1). I’m calling from a friend’s phone, mine’s broken (note 2). I’ve had an accident and I need €3,000 right now (note 3). Please don’t tell anyone.”
- note 1, “it’s me”: AI can imitate a familiar voice
- note 2, “I’m calling from a friend’s phone, mine’s broken”: An excuse for the unknown number
- note 3, “I need €3,000 right now”: Crisis and rush: hang up and check
Three steps that almost always apply
- StopHurry, fear or a big win are signals to slow down. Big decisions can wait until tomorrow.
- Check another wayHang up and call back on a number you know. Type the address yourself instead of clicking the link.
- Ask someoneShow the message to someone you trust. Anyone can fall for a scam; there is nothing to be ashamed of.
Has it already happened? Don’t wait and don’t be ashamed. Call your bank straight away (the number is on your card or in the app), then the police.
AI-assisted scams
77.5%
of respondents are worried about scams that use AI. It is the most common concern in the whole barometer.
AI helps scammers write flawless messages, imitate a voice and reach more people at once. The tricks stay the same, though: hurry, false authority, fear and secrecy.
Agree on a family code word: a word or question only your family knows. Anyone calling “from the family” and asking for money has to say it. And never send money to a “safe account”; there is no such thing.
What you can do right now
- FreeDon’t Get Scammed Four lessons: scam texts and emails, phone calls, investment scams and what to do if it happens.
- €19.90 · first lesson freeSafe and Calm Online Seven lessons in larger print, including one on voices made by artificial intelligence. It comes with the family code word and a fridge cheat sheet.
- Free toolLink X-ray Breaks the address from a text or email into parts and shows who really owns it. It opens and sends nothing.
- Free toolAI message check Paste the text of a suspicious message and a language model marks possible tricks in it. The output is labelled and can be wrong.
The limits: Neither the Link X-ray nor the AI check can tell you for certain that a message is safe. They show warning signs. You have to check and decide yourself, ideally together with someone you trust.
Deepfakes and fake content
72.3%
of respondents are worried about deepfakes and fake content. Only 26.4% believe they would recognise a convincing deepfake.
A photo, video or voice created or altered by artificial intelligence to look like a real person or a real event.
Don’t check the pixels, check the origin: who published it first, does a trustworthy source report it, do the place and time fit? For a voice on the phone: hang up and call back on a number you know.
What you can do right now
- FreeAI and Truth How language models work, how to check AI output and how to deal with deepfakes and synthetic media.
- Manipulation atlasDeepfake What a scam with an imitated voice or face looks like, and how to check it.
- Manipulation atlasAI artefacts Traces AI sometimes leaves in images, and why relying on them is not enough.
The limits: Flaws in a picture, such as garbled lettering or odd hands, only give away the poorer fakes. A good deepfake has none. That is why the origin decides, not what you see.
AI-assisted cyberattacks
70.8%
of respondents are worried about cyberattacks that use AI.
Flawless phishing emails, fake IT support, requests for text-message codes and for installing a remote-control app. At home and at work.
Never tell anyone a code from a text message, not even “the bank”. Never install a remote-control app on your phone or computer for anyone. At work: confirm a change of bank details or an unusual payment another way, and with two people.
What you can do right now
- Manipulation atlasRemote access Why you should never install an app that lets someone see and control your phone or computer.
- FreeSend a suspicious message Send us a suspicious message. We black out personal details, and the edited example can warn others.
- €60.00 · first lesson freeManipulation and Disinformation at Work Social engineering and fraud against companies: fake invoices, changed supplier bank details and deepfake calls.
The limits: We are not a security service or a helpline. We don’t watch over your accounts and we don’t reply to messages you send us. If you have already entered your details or paid, call your bank and the police straight away.
Children and AI
70.1%
of respondents are worried about how children use AI. 71.5% support teaching safe use of AI in primary and secondary schools.
Children come across fake videos, scam “prizes” in games and chatbots that sound certain even when they are wrong.
What works best is talking with children about what they see on their phones and practising together on made-up examples. The missions for children are free, need no sign-up and can be projected in class.
What you can do right now
- OverviewKids and schools Missions by age, classroom mode, and children’s safety and privacy.
- Free, no sign-upJunior Detectives Missions for ages 8–11: fact or opinion, advertising, photos that lie, privacy and chain messages.
- Free, no sign-upNot Falling for It Missions for ages 11–15: scams in games, AI making things up, edited screenshots and deepfakes.
- €19.90 · first lesson freeSmart and Calm Online A course mainly about scams aimed at parents. The last lesson covers children online and sharing their photos.
- €40.00 · first lesson freeMedia Literacy for Teachers: Ready-to-Teach Lessons Ready-made media literacy lessons for teachers.
The limits: The missions are not parental control or a content filter. Blackmail or misused photos are not a topic for a game. Talk about them with children; a child helpline can help (116 111 in many European countries).
Disinformation and manipulation
No barometer figure
The barometer did not ask about disinformation as such. Fake content made with AI is, however, its newest form.
AI has made convincing articles, fake accounts and out-of-context quotes cheap to produce. The tricks are old: emotion, “us against them”, false authority.
Before you share something: stop, check who is behind the source, find a better source and trace the original. It takes a few minutes.
What you can do right now
- FreeHow to Verify Information in 5 Minutes The SIFT method, lateral reading and tracing the origin of a story, photo or quote.
- FreeManipulation atlas Manipulation tricks with made-up examples, corrected in red pencil.
- FreeCritical Thinking Basics Fact or opinion, cognitive biases and a quick way to check information.
The limits: We don’t say who is right in political disputes. We take apart tricks, not parties.
AI and jobs
39.5%
of working respondents are worried about their job because of AI. The question went to 850 employees and self-employed people.
The barometer itself points out that the worry says nothing about the actual likelihood of losing a job. What is certain is that AI is entering the workplace, and with it errors that can sound convincing.
The more work is done with AI, the more valuable it is to judge its output: check the facts, notice when the model makes things up, and decide under uncertainty. These skills can be learned.
What you can do right now
- Certification · includes a paid course (€60.00)Critical Thinking at Work Three courses and a final test with a verifiable certificate: argumentation, decisions under uncertainty and manipulation at work.
- FreeAI and Truth The last lesson is about AI at work and at school: sensitive data, copyright and when not to use AI.
- FreeDecisions Under Uncertainty Thinking in probabilities, the outside view and meetings where dissent gets heard.
The limits: We don’t offer retraining or career advice. The courses help you work sensibly with AI; they won’t find you a new job.
What a course won’t solve
61.8% of respondents are worried about losing control over advanced AI. That is a matter for research, regulation and public debate, not for a course. What helps is at least understanding how language models work and where they fail.
First lesson: How language models work →What we do ourselves
86.4% of respondents want mandatory labelling of content made mostly by AI. We already label it: the AI check’s output is always marked, and our About page explains how we used AI to create the courses.
How we use AI →Talk it through in person
Kriticky naživo talks and workshops for libraries, senior clubs, schools, councils and companies in the Czech Republic. Using real messages with personal details blacked out, people practise spotting a scam, take home a cheat sheet and can continue online. The events are held in Czech.
Kriticky naživo