Censorship, AI and Deepfakes: The Online Battle Over the ‘Flamingo Revolution’

By: Ilda Hoxha

Since early June, activist and communications expert Alesia Burnazi has turned her Instagram account into a reporting window from the “Dëshmorët e Kombit” boulevard in Tirana.

With a mobile phone in hand and a small “fly” microphone, Burnazi began reporting on the protests known as the “Flamingo Revolution” to fill at least part of the information gap created by the traditional media boycott.

But she found a new audience on Instagram and continued her live reporting, often perched atop a concrete block or flowerpot along the boulevard to secure a fuller view of the protest.

But the attacks were not long in coming.

Burnazi was first targeted by government representatives and anonymous profiles, who accused her of manipulating videos to inflate the number of protesters, an accusation she calls ridiculous.

Then, in late August, ten protest videos were deleted from her Instagram profile under the pretext of copyright violations — and on September 3, her account was deactivated entirely without any explanation.

“This was a direct attack on freedom of expression and independent journalism,” says Burnazi.

Begun as a spontaneous protest against the government’s plans to build a massive tourist resort in the protected area of Zvërnec, the “Flamingo Revolution” quickly became the largest anti-government protest to take place in Albania since the fall of the communist regime in the early 1990s.

Yet the battle is not being waged only in the streets; it has also shifted into the virtual space of social media.

Data obtained by BIRN shows that throughout the marathon protests a whole arsenal of tools was used to silence critical voices, including hundreds of anonymous profiles, Artificial Intelligence, deepfake technology and sophisticated attacks to censor online dozens of activists, journalists, opposition figures, citizen platforms and the Instagram accounts of major media outlets.

The global organization “Repro Uncensored,” which documents digital censorship harming freedom of expression, classifies the “Flamingo Revolution” as the most visible example of efforts to silence critical voices in a society, although the malicious actors remain unidentified.

“Albania may be the clearest example we have documented, but the questions extend beyond a single country,” says Martha Dimitratou, executive director of “Repro Uncensored.”

“If coordinated actors can exploit platform systems to erase activists, journalists, major media and elected officials from public debate, that creates a model that can be replicated anywhere,” she added, calling it bad news for democracy in Albania and beyond.

AI, a double-edged sword

The “Flamingo Revolution” in Albania is unfolding in parallel with the Artificial Intelligence revolution — a new tool widely used by activists to produce creative and satirical content aimed at the old political class that has dominated the country for three decades.

But Artificial Intelligence is a double-edged sword; it was also used to spread disinformation and false information about the protesters and the protest.

Through an online search, BIRN identified dozens of cases where Artificial Intelligence was used to spread false information, which was then amplified through anonymous pages or profiles.

This is confirmed by cybersecurity expert Besmir Semanaj, who notes that AI systems have identified the “Flamingo Revolution” as one of the main developments in Albania.

“AI has been part of the online communication ecosystem around the protest, especially for creating and processing content,” Semanaj, a supporter of the protest, told BIRN.

“But a distinction must be made: using AI for communication or content creation is one thing; using AI as a weapon to censor or attack protesters is another. The protesters are using it for creativity, while the government for disinformation,” he added.

Digital policy expert Alba Brojka shares the same view. Brojka says AI has been used in both camps, but the goals were different.

“The use of Artificial Intelligence in the protest has been massive, and not only because protesters and every actor chose to use it, but it also follows the global trend,” said Brojka.

Brojka recalls cases where communication tools were used to spread fake images, known as deepfakes, to feed the narrative of foreign actors interfering during the protest.

“The photo of people claimed to be arriving by bus from Greece was a deepfake thrown online specifically to manipulate public opinion and serve a certain narrative and an argument circulated by the government, as well as by supporters of the other, non-protesting camp,” Brojka emphasized.

Behind the use of deepfakes, disinformation campaigns and digital attacks lies a deeper problem tied to the lack of legal regulation. According to Brojka, Albania lacks a proper regulatory framework to demand accountability or transparency about how these systems are used.

“At the moment we have AI use regulated only in the national digitalization strategy 2022–2026 and we don’t yet have an AI strategy,” said Brojka.

On the same issue there are also two draft laws put out for consultation, but the absence of adopted legislation makes it difficult to have transparency about how Artificial Intelligence is being used.

Digital censorship

In early September, dozens of Instagram accounts were deactivated without any warning by the company Meta, including activists, journalists, opposition figures as well as the pages of Euronews Albania and Citizens.al.

The organization “Repro Uncensored” has verified at least 70 cases — all linked to the “Flamingo Revolution.” But it believes the true scale could reach several hundred.

The Albanian government has denied any possible involvement in these efforts, and Dimitratou says that so far the investigations have found no concrete evidence implicating the government or any public institution.

Yet the facts point to coordinated targeting, with verified cases tied to the same protest, repeated copyright complaints within short periods, and replacement accounts targeted after the original accounts were deactivated.

Moreover, Dimitratou told BIRN they have obtained communications in which a reporter admits he was paid $1,800 in cryptocurrency to target several accounts, while using a phone number registered in the United States of America.

While the targeted individuals believe the campaign may be politically motivated, Dimitratou says investigations continue to uncover who ordered and funded the campaign to delete the Instagram accounts of the protest’s supporters.

Still, the concern goes beyond identifying the perpetrators and Albania’s case.

“Meta should treat these cases as a coordinated pattern and not as a collection of isolated enforcement errors, preserve all relevant data and investigate who ordered and funded the reporting activity,” says Dimitratou.

“Without transparency, meaningful safeguards and accountability, platform infrastructure can become a ready-made tool to crush political opposition, independent reporting and democratic participation,” she further emphasized.

In the same vein, Semanaj also speaks of coordinated targeting in the case of the Instagram account deactivations, with the aim of silencing critical voices in support of the “Flamingo Revolution.”

Beyond the technical side, Semanaj also notes the deep contrast in how state institutions react when the government’s interests are hit versus cases where critical voices are targeted.

“Another concerning thing is the government’s double standard. When businesses close to it were hit by comments on Google, the government mobilized every institution and contacted Google. For the case of the protesters’ accounts, nothing was done. It simply stated through Rama that ‘we didn’t do it,'” he emphasized.