In the mid-2020s, AI-driven disinformation has become a new and quiet threat to internal security. Deepfake videos, fake images, and AI-generated texts can target societies, especially during crises, and create chaos. Unlike traditional propaganda or simple fake news, this new type of disinformation can produce more believable content and spread it very fast. Because of this, students who study international relations, hybrid threats, and internal security are paying more attention to the topic. During a crisis, a deepfake speech of a national leader, a fake official document, or a manipulated emergency alert can cause panic and distrust when society is already fragile.

As German Foreign Minister Annalena Baerbock said, “Artificial intelligence makes disinformation cheaper, easier, and more effective.” And she is right. With just a normal computer and an internet connection, people can now create false content that only professional intelligence teams could produce in the past. This makes the asymmetric threat even stronger. A small group, or even one person, can use AI tools to mislead large populations or put state institutions in a difficult situation. These actions usually cost very little, but the damage they can cause can be extremely large.

Examples of deep fakes in times of crisis

  1. In a time of war and invasion: in the first months of the 2022 invasion of Ukraine, the world saw maybe the first “weaponized” deepfake made against a state leader. In March 2022, a fake video appeared on social media that tried to copy Ukrainian President Volodymyr Zelensky. In the video, Zelensky’s face and voice were recreated with AI, and it looked like he was telling the Ukrainian people to “lay down their weapons” and surrender to Russia.

    This deepfake even slipped into the broadcast of one of Ukraine’s official TV channels for a short time, with a subtitle bar under it. The goal was clear: to break the morale of the Ukrainian people during the war, to weaken the soldiers’ resistance, and to push them toward surrender. The Ukrainian side probably expected such a threat, because right after the video appeared, Zelensky made a real statement explaining that it was a provocation. Platforms like Facebook, YouTube, and Twitter also reacted fast and removed it. In the end, the deepfake of “Zelensky calling for surrender” did not survive for long. It looked a bit artificial from a technical angle, and because Ukraine reacted quickly, its effect stayed limited. But this incident was still a warning shot. If the deepfake had been made with higher quality and if it had not been disproved so fast, the results could have been much more serious.

    Imagine soldiers in the middle of a war suddenly seeing their leader on TV raising the white flag. Even a few moments of doubt or panic could cause collapses at the front line. Right after the Zelensky video, another deepfake started to spread online, this time as a kind of “retaliation”: a fake Putin telling Russian soldiers, “Surrender, go home.” Both cases showed that deepfake videos have already become a new psychological warfare tool during conflicts.

  2. In terror and emergency situations: we already know from past events how dangerous false information can be during a major terror attack or a national emergency. In 2013, hackers took over the Associated Press Twitter account and posted a fake tweet saying there was an explosion at the White House. This caused sudden panic and even shook the financial markets for a short time. Today, with artificial intelligence, it is possible to create fake warnings that look much more realistic.

    For example, in early 2024, a video spread quickly on social media in the United States. It showed what looked like nuclear attack alerts in several states. The video seemed to be recorded from the Emergency Alert System on TV, and it said that some big cities needed to be evacuated. Of course, it was not real. It was actually a fictional scenario video uploaded to YouTube two years earlier. But when someone shared it as if it were happening now, it scared thousands of people. FEMA had to make a statement immediately saying, “We did not issue such an alert.” This example shows that even emergency systems can be abused with AI-powered disinformation.

    Even worse is the possibility that bad actors might launch coordinated disinformation attacks during an actual terror event. Imagine a big bombing happens, and right after it, fake accounts and AI bots start posting. They spread fake images, false reports of more attacks, or claims blaming a certain ethnic group. Society is already in panic at that moment, and such provocative disinformation can create a second wave of crisis. Security forces may be misled, and people may not understand what is real and what is not. In some cases, this could even lead to mob violence or social clashes. The fear that terrorism wants to create becomes even stronger because of AI.

  3. During elections and periods of political chaos: in democracies, election periods are already times when misinformation spreads easily. But artificial intelligence can make this situation much worse. The 2023 elections in Türkiye showed a clear example of this. Just a few days before the vote, one of the presidential candidates, Muharrem İnce, had to withdraw because of a video that appeared online. The clip showed him in an inappropriate situation, and many people claimed it was a deepfake shared to damage his reputation. İnce said the video was fake and edited, and that it hurt his political career, so he decided to leave the race. In the same election, another incident happened.

    At government rallies, a montage video was shown that placed the opposition leader together with members of a terrorist group. President Erdoğan even played this video on stage to suggest that the opposition was cooperating with the terrorist group PKK. Later, it became clear that the video was not real and that the images were edited. These kinds of political disinformation attempts make polarization deeper and put a shadow over the election. When voters cannot understand if a video is real or fake, their trust in the democratic process becomes weaker.

    On top of that, AI-powered bot accounts and fake content can create a kind of “noise wall” on social media, making it hard for real news to reach people. During the 2023 elections, the opposition openly accused Russia of running a “deepfake campaign” against the country. Whether this claim is true or not, the important point is clear: a foreign actor can now use AI technology to spread disinformation during an election, and they can do it secretly and at low cost. This has become a serious security concern for all democratic states, including NATO allies.

  4. Fake documents and official letters: disinformation does not only come in the form of videos or images. It can also appear as fake texts and documents. With AI-supported language tools, it has become easier to create fake news articles, press releases, or government orders that look convincing. For example, during a crisis, a fake directive that seems to come from an official institution could mislead people.

    During the Russia-Ukraine war in 2022, a false letter about Poland spread on social media. It claimed that the Polish government was planning to send Ukrainian refugees in the country to the front by force. This “document” was completely fabricated, but pro-Russian accounts shared it to create distrust between two friendly nations. Similar tricks include fake statements written as if they were from NATO officials or false reports presented as leaked EU documents. These types of disinformation tools are now part of the wider toolbox. Fake military or political documents can be especially harmful if they appear during a crisis. Decision-making systems inside the state can be disrupted.

    Security institutions might spend time dealing with a fake order while responding too late to a real threat. Misunderstandings can also grow between allies. The examples in this section show how AI-powered disinformation is used in war, terrorism, elections, and other crisis situations. Deepfake videos, fake images, and false documents are activated exactly when societies are most vulnerable, creating an asymmetric impact. So the next question is: why is this threat seen as “asymmetric,” and how is it discussed in transatlantic security circles?

Asymmetric threat and the transatlantic security dimension

AI-driven disinformation is called an asymmetric attack method because it has the potential to break traditional power balances. In the past, to threaten a country’s internal security, an enemy needed expensive and risky operations—like sending agents, spreading propaganda on the ground, or planning terror acts. But now, someone behind a keyboard, maybe even a person with no clear identity or a bot network, can create chaos with a deepfake video or a fake news article.

Asymmetric means that the weaker side can harm the stronger one in an unexpected way. A small disinformation operation can work almost like a digital “Trojan horse,” creating problems from the inside. NATO and its allies have put this topic on their agenda, especially because of Russia’s hybrid warfare tactics in recent years. In NATO’s 2022 Strategic Concept, members are encouraged to improve their resilience against hybrid threats, including cyberattacks and disinformation campaigns. Some states, especially Russia, use information warfare methods when their conventional military power is not enough.

Analyses from the NATO Defense College say that in a crisis, Russia might try to damage the reputation of leaders with deepfake videos or confuse decision-making processes with fake news. One article even warned that Russia could spread disinformation right after a terror attack to create fear and weaken trust in the government. In such a scenario, imagine a bombing happens in a NATO country, and immediately there are false reports of a “second bomb,” fake warnings supposedly from the health minister, or lies blaming certain groups. While people are unsure what to believe, security forces may follow fake leads. This reduces public trust and makes crisis management harder. In the transatlantic security framework, the idea of collective defense is expanding to include protection against hybrid attacks. NATO has StratCom (Strategic Communications) teams and Cyber Security Centers working on detecting and preventing AI-powered disinformation.

In 2023, NATO allies tested deepfake-detection systems in joint exercises for the first time. The EU and the US are also cooperating with social media platforms to develop faster ways to flag and remove fake content. International cooperation is essential because disinformation does not respect borders. A false story released in one country can spread across the world within seconds. Rumors that could harm unity inside NATO—like “Country X will not support Country Y”—must be corrected quickly, or they can weaken trust within the alliance. It is notable that organizations like NATO now treat the fight against disinformation as seriously as traditional defense.

The idea of democratic resilience is becoming more important: societies must be able to resist attacks, not only from armies. Countries like Finland and Sweden have been teaching media literacy in schools for many years, helping citizens become more resistant to false information. This builds a culture where people check official sources first, instead of believing everything they see during a crisis. Inside the transatlantic community, such good practices are being shared more and more.

Of course, the issue is not only about defense. There is also a deterrence side. NATO experts argue that hostile states should pay a price if they use these methods. For example, if it is proven that state-supported groups in a country carried out deepfake attacks against another society, diplomatic or economic sanctions could be considered. The goal here is to treat disinformation as a type of attack and to show that there will be a response. To sum up, AI-powered disinformation is taken seriously in the transatlantic space and is becoming part of collective security strategies. NATO and its allies are trying to build both technological and social resilience against it as part of hybrid warfare. But how effective these efforts will be also depends on how well individual countries prepare themselves. And this brings us to the next point: what can be done at the national level, and what might happen if nothing is done?

Conclusion and warnings

In the age of artificial intelligence, fighting disinformation has to become an essential part of internal security policies. The deepfake videos, fake crisis images, and manipulative content we saw in the examples above will appear more often and with stronger impact in the future if proper measures are not taken. So what should be done?

First, state institutions and media organizations need a shared system for quick and reliable communication during a crisis. If a fake video is spreading, the relevant leader or institution should have a direct way to deny it immediately. In the Zelensky example, the president stopped a major disaster simply by going in front of the public quickly and saying, “We are not surrendering; the video is fake.” Every country’s leadership should be able to react like this. If the response comes too late, fact-checking might not work; the false story may already be fixed in people’s minds.

Second, better cooperation with technology companies and social media platforms is necessary. As deepfake-detection algorithms improve, they need to be used on social networks. But this is a race: as AI fakery becomes more advanced, detecting it becomes harder. If platforms do not take this responsibility seriously, the information space of societies will keep getting poisoned. For example, the fake Pentagon explosion image spread very fast on Twitter from accounts with blue check marks and even affected financial markets for a short time. This happened because of weaknesses in the platform’s verification system. If social media companies do not take stronger steps in identity verification and content moderation, the next fake crisis might lead not only to economic losses but also to real casualties—for example, a crowd panic where people get injured.

The third important point is public awareness. Media literacy and disinformation awareness should be included in school programs and public campaigns. People need to learn to be careful, especially in crisis situations, when they see shocking claims online. If society does not develop a habit of “don’t trust the first thing you see; check the source,” then disinformation will succeed no matter how advanced the technology becomes. In this sense, we need a balance between a “culture of doubt” and avoiding paranoia. People should not believe everything, but they also should not think every real news story is fake. This balance is difficult but possible.

Finally, legal and criminal regulations need to be updated. Someone who creates a deepfake video during an election or spreads a fake emergency alert might not fit clearly into current criminal law, because the technology is new. Countries should update their laws to deter disinformation attacks and work together internationally to track these crimes. Otherwise, bad actors will continue without consequences. In the long term, creating international norms against state-sponsored disinformation campaigns may also become necessary.

If we are not proactive in the fight against digital disinformation, the vulnerable moments of societies will remain as security weaknesses. War, terror attacks, natural disasters, or political crises… In exactly these moments, AI-generated fake content can show disasters that are not real, create chaos, or make an existing crisis deeper and harder to manage. If we need to end with a clear warning of “what happens if we don’t act,” it can be said like this:

  • If states and technology platforms do not develop systems to detect deepfakes and AI disinformation now, then in the next crisis a fake leader’s speech could push armies to follow a wrong order, a fake alert could make whole cities evacuate, or a false news story could turn people against each other.

  • If the international community does not take deterrent steps against the actors who spread disinformation, authoritarian regimes and terrorist groups will use this cheap tool even more, and they will not hesitate to target the internal stability of democratic countries.

  • If we as individuals do not learn to look critically at the information we consume, then all of us can become victims of disinformation—people rushing to the banks because of a false tweet, crowds filling the streets because of a fake video, or even citizens thinking their neighbors are enemies.

Let’s remember that internal security is not only about physical threats inside our borders. We also need to protect our minds and our social unity. AI-powered disinformation targets exactly these things. If we do not take precautions today, it is very possible that tomorrow we face a kind of chaos where the line between truth and lies disappears completely. Because of this, building resilience before crises happen and defending the truth calmly when they do happen should become the new normal for both individuals and states.