The Bulletin of the Atomic Scientists set the Doomsday Clock to 85 seconds to midnight, which is the closest we have ever been to the ‘end of the world’ in history. While this measure of how far the world is to ‘the end’ is not guaranteed, with the rise of geopolitical conflict, climate instability, biological threats, nuclear risks, and disinformation of shared reality, these pressures push humanity closer to an existential point of global annihilation. Significantly, however, disinformation in the context of Doomsday functions as a multiplier of existential threat by corrupting the shared reality needed for NATO allies to act as a unified force. Unlike disinformation tactics from non-allies, disinformation between and within NATO allies creates internal friction that erodes the core foundation of the alliance: collective trust. If allies are deceived into seeing their national security spending as a drain rather than a necessary investment in safety, they can be less willing to support one another. That reluctance weakens NATO’s ability to respond effectively if Article 5 is triggered by an attack.
AI-generated disinformation has significantly escalated fears and distrust during the recent 2026 tensions between the U.S. and Iran. This vulnerability becomes even more serious as mistrust between these states is already high, having a history of confrontation and broken commitments. This phenomenon of AI-generated disinformation is often referred to as the “digital fog of war,” which has created an environment where neither citizens nor military analysts can reliably distinguish between authentic conflict and machine-made propaganda. There has been a significant uptick in the Iranian regime targeting American sources such as government agencies, infrastructure, and companies. Canada is also being targeted as well, especially because we share intelligence and media ecosystems with the U.S. Although we are not militarily involved in this conflict, because we have expressed support for the U.S., Iran now sees Canada as another target to digitally exploit. In moments like these, AI-generated disinformation can quickly spread before governments can confirm and debunk misunderstandings. This can create feedback loops among Canadian consumers, as Canada can unintentionally reinforce U.S. misinterpretations, since both countries share a highly integrated and borderless media ecosystem, which can strain trust between these NATO allies.
Previous incidents illustrate just how rapidly AI-generated disinformation can spread, causing significant effects before authorities are able to conduct verification. In 2023, an explosion at the Pentagon found its way to news networks, which caused the stock market to dip. Someone had posted an AI-generated image of an explosion at the Pentagon, using an account that was linked to the Bloomberg news organization on Twitter, which then got shared across multiple other verified accounts on other platforms. The Dow Jones Industrial Average fell about 80 points within a four-minute window and the S&P 500 slipped into negative territory during the same time frame. The image spread widely before officials could confirm this incident had not occurred. While this appeared to have been coordinated, experts still cannot trace down the suspect. The Pentagon hoax demonstrates how a single AI-generated image can deter financial markets in minutes, which can create a similar shock that easily fractures the diplomatic trust and rapid decision-making required for collective defence during a geopolitical crisis.
Modern news outlets and social media prioritize speed over verification. This does not suggest that journalists are careless; rather, advancing tactics of misinformation expose the structural flaws and exploit the economic pressures built into these contemporary media systems. Platformers, anchors, and broadcasters are incentivized to be the first to report because outputting early can attract more views, clicks, and engagement, which often generates more advertising revenue and reliability. However, this defeats the purpose of journalism, which is to establish a fundamental factual baseline; because of rapid technological advancement in the information space, policies used to prevent the dissemination of misinformation in the news need to keep pace with the increasingly changing atmosphere. The race for speed over certainty doesn’t just distort the news cycle, it creates the perfect conditions for AI-misinformation to spread like wildfire across borders and unsettle the shared reality among NATO allies.
AI-generated misinformation is not exclusively limited to images or written texts; it also appears as audio as well. In a viral clip with a voice resembling Trump, he appeared to be threatening Justin Trudeau with arrest in a heroic-sounding tone, making the president seem like a saviour—rescuing Canadians out of their misery. This audio clip was spread across different social media platforms in November 2024, gaining millions of views, but was ultimately debunked as a deepfake spoof fabricated to create a false impression of hostility between the two close allies. Although governments did not treat the viral audio specifically as a diplomatic threat, the case raises several questions about who exactly people can trust when truth online increasingly becomes harder to distinguish and how we can manage an increasing friendly-fire of misinformation, especially as this audio clip spread during the lead-up to and immediate aftermath of the U.S. election, tapping into partisan anxieties in both countries. Many users, out of meme culture, shared this clip as a joke or piece of satire, while others reposted it as factual news, causing widespread confusion. Another piece of context to consider is that during the time the audio was gaining virality, Trump publicly threatened to impose 25% tariffs on Canada, which made his tone in the deepfake seemingly plausible to some listeners. The AI-generated audio showcases how easily a synthetic clip can ignite confusion across borders, creating a friendly-fire of misinformation that spreads faster than allied institutions can contain it.
During the early days of deepfakes, humans were quite good at spotting AI. However, due to the sheer volume and speed of synthetic media now, humans cannot reliably distinguish output from a real person, which has shifted us to employing an AI-on-AI verification. This creates a “cat and mouse” arms race where, although detection algorithms improve spotting fakes, AI simultaneously learns how to produce even more human-like content, increasing the potential for AI detection false positives. For allies like Canada and the U.S., this “friendly-fire” of disinformation is now treated as a hybrid threat. To counter this, Canada has established the G7 Rapid Response Mechanism, a collaborative initiative designed to identify and respond to foreign threats to democracy, focusing on disinformation and foreign information manipulation and interference. While these institutional defences provide a necessary shield, as AI continues to outpace policies and human detection, they leave us to navigate a future where the line between a verified ally and a synthetic adversary becomes increasingly indistinguishable.
Photo: “Photo of Laptops on a Wooden Table” (2021), by Ömer Tosun via Pexels. Licensed under the Pexels License.
Disclaimer: Any views or opinions expressed in articles are solely those of the authors and do not necessarily represent the views of the NATO Association of Canada.
