MarGib – The World Behind the Keyboard: Deepfakes: The New Era of Disinformation – How Artificial Intelligence Manipulates Reality

Deepfakes: The New Era of Disinformation – How Artificial Intelligence Manipulates Reality

MarGib August 19, 2026

🌐 🇵🇱 Polski · 🇬🇧 EN

Deepfakes are becoming increasingly realistic, and their use for disinformation purposes is growing at an alarming rate. From fake statements by politicians to financial fraud – artificial intelligence is changing the rules of the game. How can we defend against it?

Ilustracja porównująca prawdziwe wystąpienie polityka z jego deepfake’ową wersją, z lupą wskazującą na różnice. — Deepfake vs. reality: How artificial intelligence distorts the truth.

In 2024, deepfakes ceased to be a technological curiosity – they became a tool for disinformation, fraud, and large-scale manipulation. Artificial intelligence generates fake videos, voices, and images with such precision that even experts struggle to identify them. Are we prepared for this new reality?

Deepfakes in action: The most high-profile cases of recent months

Over the past year, deepfakes have made headlines due to several spectacular abuses. In February 2024, in the USA, scammers used a fake voice of President Joe Biden to urge voters to boycott the primaries. The FBI confirmed that the matter is being investigated as an attempt to interfere with the democratic process. Similar incidents occurred in Hong Kong, where a company lost $25 million after employees received fake transfer instructions from a supposed CFO – his voice and image were generated by AI.

In Poland, deepfakes have appeared mainly in a political context. In March 2024, a video of Prime Minister Donald Tusk allegedly announcing tax hikes spread across social media. Although the material was quickly debunked, it managed to cause confusion and online debate. This shows how easily fake content can influence public opinion.

Not all cases are so obvious. In 2023, during the parliamentary election campaign, a deepfake of an opposition party leader suggesting corruption was detected. The material was removed by social media platforms, but its impact on voters remains difficult to estimate. Artificial intelligence doesn't ask if we are ready – it is already operating, and we must learn to live with it.

How are deepfakes created? Tools and algorithms that change the game

Creating deepfakes has become easier than ever. Thanks to open-source tools and generative models, even people without advanced technical knowledge can generate realistic fakes. Here are the most popular solutions:

deepfacelab – open-source software for face swapping, used in over 90% of video deepfakes. It requires basic technical knowledge, but there are tutorials available that explain the process step-by-step.
Stable Diffusion – a text-to-image model that allows for generating realistic photos based on descriptions. It can be used to create fake evidence or compromising materials.
elevenlabs – a voice cloning tool that enables the generation of speech that sounds like real recordings. Just a few seconds of an original voice are enough to create a faithful copy.
DALL·E 3 – another model from openai that generates images based on text. Although the company has introduced restrictions to prevent abuse, the tool can still be used in an unethical way.

Moreover, "deepfake-as-a-service" offerings are appearing on the market, providing custom fake content generation for just a few dollars. The Bible of modern AI contains more information about tools and technologies that can be used for both good and bad purposes.

How to detect deepfakes? Methods and their limitations

In the fight against deepfakes, detection tools are becoming crucial. Technology companies and research institutions are working on algorithms that can identify fake content. Here are the most important solutions:

Microsoft Video Authenticator – analyzes artifacts in the image, such as unnatural movements or blurs, which may indicate manipulation.
Deepware Scanner – an open-source tool that scans videos for traces of AI generation.
DARPA medifor project – a research program funded by the US Department of Defense that focuses on detecting media manipulation.

Unfortunately, these methods have their limitations. Studies show that even the best algorithms have an effectiveness of around 90%, but only under controlled conditions. The problem is the so-called "arms race" – deepfake creators quickly adapt to new detection methods, making the fight against fakes an endless chase.

It is also worth remembering social initiatives such as invid-WeVerify or Truepic, which offer tools for verifying content authenticity. Truepic uses blockchain technology to certify photos and videos, which can be an effective way to combat disinformation.

Legal regulations: Is the law keeping up with technology?

Deepfakes pose new challenges for lawmakers. In the European Union and the United States, attempts are being made to regulate this issue, but the results are mixed.

European Union

Digital Services Act (DSA) – imposes an obligation on social media platforms to label and remove harmful AI-generated content.
AI Act – classifies deepfakes as "high risk" and requires labeling of AI-generated content. The act will come into force in 2026.

United States

DEEPFAKES Accountability Act – a draft bill that requires labeling of deepfakes and provides penalties for their malicious use.
State of California – prohibits the use of deepfakes in political campaigns within 60 days of an election.

Poland

In Poland, there is no dedicated law regarding deepfakes yet. Abuses can be prosecuted under the provisions of the Penal Code, such as defamation or violation of personal rights. However, the lack of specific regulations makes fighting deepfakes more difficult.

Why are deepfakes so effective? Psychological mechanisms

Deepfakes work because they exploit the natural tendencies of the human brain. Here are a few key mechanisms:

"Illusory truth" effect – people are more likely to believe information that is repeated, even if it is false.
Perceptual realism – the brain has a tendency to accept realistic images as true, even if they are fake.
Confirmation bias – deepfakes are more effective when they confirm the existing beliefs of the audience.

Studies show that even people aware of the existence of deepfakes can fall victim to them. An experiment conducted by the University of Amsterdam showed that 60% of participants believed a deepfake of a politician if it aligned with their views. In turn, a study by Stanford University showed that voice deepfakes are harder to detect than video – 73% of respondents did not recognize the fake recording.

Which sectors are most exposed to deepfakes?

Deepfakes can be used in various sectors, but some are particularly vulnerable:

Politics – deepfakes are used for election disinformation, undermining trust in institutions, and manipulating public opinion.
Finance – scammers use deepfakes to extort money by impersonating company executives or well-known personalities.
Social media – fake content spreads quickly, affecting users' perception of reality.
Cybersecurity – voice deepfakes are used in phishing attacks, e.g., impersonating a company CEO.

It is worth noting that deepfakes can also have positive applications, e.g., in education or entertainment. However, it is precisely their potential for harm that makes them so dangerous.

How is Poland fighting deepfakes? Educational and technological initiatives

In Poland, various initiatives are being undertaken to counteract disinformation using AI. Here are the most important ones:

Government initiatives

"Digital Poland" program – as part of the project, NASK conducts research on deepfake detection and public education.
"Sprawdzam" (I verify) campaign – an initiative of the Ministry of Digital Affairs and govtech Polska, which promotes information verification.

Non-governmental organizations

Panoptykon Foundation – conducts workshops on digital hygiene and critical thinking.
Demagog – a fact-checking organization that monitors deepfakes in the Polish public sphere.

Technological solutions

Warsaw University of Technology – work on algorithms detecting deepfakes as part of the AI4Media project.
neurosys company – develops tools for analyzing the authenticity of video content.

Despite these actions, Poland still lags behind countries like the USA or EU nations in terms of regulations and education regarding deepfakes. Further investment in detection technologies and public awareness is needed.

What can you do to avoid being deceived?

Deepfakes are becoming harder to recognize, but there are ways to reduce the risk of falling victim to them:

Check sources – always verify where the content comes from. Is it published by trusted media?
Look for inconsistencies – pay attention to unnatural movements, blurs, or sound discrepancies.
Use verification tools – such as InVID or Truepic, which help check the authenticity of content.
Be skeptical – if something seems too good (or bad) to be true, it probably is.
Educate yourself – follow initiatives such as the "Sprawdzam" campaign or workshops by the Panoptykon Foundation.

In the face of the growing number of deepfakes, digital hygiene is becoming crucial. Artificial intelligence in the labor market is changing not only how we work but also how we consume information. We must learn to take a critical approach to the content that reaches us.

Summary: Deepfakes are not science fiction, they are our reality

Deepfakes have ceased to be a futuristic threat – they have become part of our everyday life. From fake statements by politicians to financial fraud, artificial intelligence opens up new possibilities for manipulation. Although there are tools for detecting deepfakes and legal regulations to counteract them, fighting them is difficult and requires the commitment of both institutions and each of us.

Building awareness and education is key. We must learn to take a critical approach to the content we consume and use available verification tools. Deepfakes will not disappear – on the contrary, they will become increasingly realistic. It is up to us whether we allow them to manipulate our reality.

"Artificial intelligence is not good or evil. It is a tool that can be used for different purposes. It is up to us how we use it."

— Fei-Fei Li, Co-Director of the Stanford Institute for Human-Centered Artificial Intelligence