deepfakes

“So what?” Richard Martin, a US-based pan-Africanist, shot back on LinkedIn, dismissing another user who dared to call a trending speech of Burkina Faso’s military leader, Captain Ibrahim Traoré, by its real name: a deepfake.

It was the kind of fiery anti-imperialist speech delivered by Muammar Gaddafi at the United Nations in 2009. 

“You ask, ‘Why is Africa poor?’ That’s the wrong question,” Traoré tells the big news companies in crisp, accented English. “The real question is, how is Africa kept poor when it’s so rich?”

Patti Boulaye, a British-Nigerian singer and actress, shared it on LinkedIn, lending it an aura of authenticity. Many of the over 300 comments under her post on the social media platform echoed her conviction that the speech was real. Another sizable fraction recognised it as fake but embraced it anyway. Among those who raised concerns was London-based IT training consultant Andrian Moore, who was quickly pushed back by Richard: So what?

Undeterred by criticism, Patti, who, according to her LinkedIn bio, leads a fundraising effort to build healthcare clinics and a school in some African countries, posted yet another AI-generated speech.

“All I ask is that you do not throw away the baby with the bath water,” she wrote.

The new video shows the same man in his signature brownish camouflage, speaking on a global podium. Like the first, which was reposted over 250 times and received reactions from more than 1,300 users, the second video also went viral.

Captain Traoré seized power in Burkina Faso on September 30, 2022, citing military leader Paul-Henri Damiba’s failure to curb an escalating Islamic insurgency, and pledged to restore democracy within two years. But as the deadline approached, he extended his rule by five more years, cementing his place as president of the Francophone nation.

In office, Traoré has overseen notable agricultural and industrial initiatives, while also drawing international criticism. In May, Burkina Faso’s military was accused of killing at least 130 civilians in Solenzo, a town in the country’s west.

His rule has been marked by bold geopolitical shifts: withdrawing from the Economic Community of West African States (ECOWAS), forging a new alliance with other military-led neighbours, severing ties with France, and deepening relations with Russia.

A dictator loved by some and criticised by others across borders, Traoré’s complex personality made him a tool of choice in the deepfake videos now making the rounds in Africa.

Many of them hit the internet in May, attributed to Traoré and conveying similar messages – Africa, a historical victim of Western exploitation, must deliver itself. 

The videos appear to premiere on YouTube before spreading to social media platforms like TikTok, Facebook, and LinkedIn. The responses, from thousands of people, are often the same: tears, admiration, accolades, as well as scepticism. 

“This is not a simple speech, it’s rather a textbook of freedom for Africa,” one YouTube comment reads.

The numerous fact-checks that followed, proving the videos are synthetic, could not stop the people from believing them.

One, “President Ibrahim Traoré’s Bold Speech to the IMF Shocks the West,” has got close to 1.5 million YouTube views, while “Africa Will Not Kneel: Traoré’s UN Speech Shocks the West and Defies Neocolonial Power” has over 600,000. Now deleted by YouTube, “Pope Leo XIV Responds to Captain Ibrahim Traoré| A Message of Truth, Justice & Reconciliation” was streamed nearly a million times. Its description as “a work of fiction inspired by the life of IBRAHIM TRAORÉ” had little effect on the thousands who flooded the comment section.

“They sound like what leaders should say but seldom dare to,” Ghanaian strategic communicator Rifkin Dodo said of the moving speeches.

Yaw Kissi, a pan-African writer popular on LinkedIn, shares the same view: “When people hear voices, even artificial ones, boldly articulating what they’ve always felt but rarely heard echoed, it sparks something powerful.”

Where are they from?

Because of their anti-West rhetoric and pan-African posture, the biggest bet would be that these speeches originated in Africa itself. Yet an in-depth analysis of seven of them, collected from different YouTube channels, tells a different story.

The earliest uploads can be traced to three channels: Pan-African Dreams, Black Rebellion, and Univers Inspirant. The latter two were both created in March this year. Though Pan-African Dreams, which premiered the viral deepfake of the newly-inaugurated head of the Catholic Church, Pope Leo XIV, responding to Traoré, has since been deleted by YouTube, archived records show it shared the same “Welcome to …” opening line in its page description as the other two.

Labelled D1 through D7, metadata extracted from the three channels was subjected to forensic linguistic analysis, using advanced AI language models.

An analysis of authorship and voice, lexical patterns, and stylistic markers revealed overlaps suggesting the channels are from one source:

Black Rebellion and Univers Inspirant both joined YouTube…in March 2025, within 9 days of each other.

Pan-African Dreams shares the same “Welcome to …” channel intro formula despite deletion.

All seven videos were released within 10 days (May 15–24, 2025) — a compressed schedule suggesting a campaign roll-out rather than organic posting.

While Black Rebellion chose a different core style, Pan-African Dreams and Univers Inspirant share the same linguistic fingerprint in their video descriptions. In one instance, Black Rebellion merged the styles of the other two channels.

This pattern implies one production team testing multiple rhetorical frames (religious/spiritual, militant/anti-colonial, media-critique) to maximise resonance across audiences.

YouTube transcripts of two videos from Black Rebellion and Univers Inspirant were also subjected to a similar analysis:

The core linguistic fingerprint—the ideology, the confrontational pronoun strategy, the complete lack of filler words, the passionate tone, and the reliance on specific rhetorical patterns—is remarkably consistent across both texts.

The significant differences in structure, pacing, and lexical focus are best explained not by a change in author but by a deliberate and skilful adaptation to audience and medium.

The underlying idiolect is the same; only the presentational style has been modified to fit the specific context.

Surprisingly, Black Rebellion and Univers Inspirant are both located in France, prompting an AI model to conclude:

The timing, registration, style overlaps, disclaimers, and cross-channel duplication all strongly suggest that Pan-African Dreams, Black Rebellion, and Univers Inspirant are not independent outlets but rather part of a single coordinated network, seeded from France in March 2025, with the May videos representing a first major campaign wave.

Truths, lies, and half-truths 

Richard repeated the same statement many times under Patti’s post: It’s absolutely true. Like him, many embrace the AI speeches, convinced they contain undiluted truths about Africa. 

But a fact-check revealed a mix of accurate data, exaggerations, and popular opinions.

Take Captain Traoré’s age. Official records show he was born in 1988, making him 37 in 2025. Yet in one of the videos, he declares: “I am 34 years old.”

Misleading claims of 46 US military bases in Africa, and the unverifiable annual transfer of $500 billion to France by the African CFA franc bloc, among others, further erode the speeches’ credibility. 

But one cannot deny the compelling manner in which facts are presented, shining light on historical exploitation and suppression of pan-African voices. 

It’s true, for instance, that Africa is a “net creditor of the world,” receiving $134 billion in aid and losing $192 billion annually through illicit channels. It’s also true that citizens derive little benefit from resources taken from African soils.

“Africa has 70 per cent of the world’s cobalt,” one deepfake Traoré declares. The mineral powers cell phones, computers, and electric cars, but many Congolese, from whose soil it is mined, cannot afford these products.

A significant share of the world’s diamonds and gold comes from Africa. But several miners in those regions remain extremely poor.

“Africa does not owe you,” Traoré tells the International Monetary Fund (IMF) in another deepfake. “You owe Africa. You owe us for the gold taken under colonial boots. You owe us for the minerals that power your smartphones while our villages remain in darkness.”

Declassified documents and government inquiries have linked Belgium to the removal of Congo’s anti-imperialism leader Patrice Lumumba in 1960, which culminated in his murder. Thomas Sankara’s 1987 overthrow and assassination have also raised questions about France’s role.

“Lumumba spoke the truth and paid the price,” said Rifnikin. “Sankara challenged the system, and bullets followed. Leaders today remember these as if it was yesterday.

“AI speaks openly because it has nothing to lose and cannot be arrested, poisoned or assassinated. African leaders remain silent because they have everything to lose: power, comfort, and sometimes even their lives.”

Yet, as Yaw warns, the same technology that gives voice to suppressed truths also carries the risk of distortion when cloaked in the aura of charismatic leaders.

“Communities must be equipped to critically engage with these narratives, separating emotional resonance from factual accuracy, while leveraging the technology to reclaim our own stories,” he said.

To what end?

“What are the objectives of those who posted the video?” Adrian asked during our conversation. “Are we advancing this agenda by sharing it?” 

A closer look at the YouTube channels provided some answers. Mid-roll and interactive overlay ads, features tied to AdSense, Google’s monetisation programme, were embedded in the videos, suggesting that profit, not ideology, is at play.

While a ‘Join’ button and fan-funding features that could have confirmed channel monetisation are absent, further analysis indicates significant background, voice, and musical variations in videos that are substantially the same. This could be explained by the YouTube policy requiring that borrowed content be significantly altered to qualify for monetisation.

Black Rebellion (28,000+ subscribers and more than 2.5 million views) and Univers Inspirant (46,000+ subscribers and more than 6 million views) are likely generating revenue from disinformation targeting Africans. On Black Rebellion, only three of its 40 videos are not about Traoré, but they are still about Africa. Univers Inspirant hosts 112 videos; just six do not feature Traoré. Those six, produced in French, highlight the achievements of France’s past and present leaders, a divergence that hints at the channel’s possible origin.

While metadata alone does not conclusively prove that the channels were registered in France, as creators can obscure their real location, the French-language content, French name (Inspiring Universe in English), and page description in French reinforce the suspicion.

“Today, we have an inspiring message from Ibrahim Traoré,” said Adrian, who also told me his surprise at finding such a deepfake circulating on LinkedIn, shared by someone from whom one would ordinarily expect due diligence.

“Tomorrow, we could have Ibrahim Traoré promoting prejudice or justifying crimes, and it would be accepted, even acted upon, simply because it came in the voice of a visionary African leader.”

Africa has endured centuries of exploitation. Yet the ongoing weaponisation of her pain, packaged through deepfakes by individuals linked to a country that once colonised her, opens a new chapter in the history of human exploitation.

“Colonialism never ended,” says Traoré in one such deepfake. “It just changed its face.”

Celebrities are all too familiar with the world of deepfakes, the colloquial term for artificial intelligence-generated videos that depict actors and other Hollywood talent falsely doing or saying things that they never agreed to.

To protect themselves, actors including Steve Harvey, Beverly Hills talent agency WME and studios have enlisted the help of Vermillio, a Chicago-based company that tracks famous people’s digital likenesses and intellectual property online. Depending on what its clients want, it can have the material taken down .

But as AI technology continues to improve and becomes more widely available to the general public, regular people are getting scammed too.

Now, Vermillio says it is offering a version of its service for free to everyone.

The move comes as more and more convincing deepfakes continue to proliferate online, making it difficult for social media sites to police such activity. In 2019, there were about 18,000 deepfakes globally and this year, there have been roughly 2 trillion generative creations, said Vermillio Chief Executive and co-founder Dan Neely.

That leaves average Joes at a growing risk of being impersonated online, with little recourse.

“We can’t wait for governments to solve this problem,” Neely said. “We can’t wait for legislators to solve this problem. We can’t wait for other people to solve this problem. We just said it’s the right thing to do, so we should just be doing it.”

With this move, Vermillo is adopting a classic “freemium” model — offering partial service for no charge and up-selling for additional features.

Here’s how it works.

Using its TraceID technology, the company flags problematic content. For paying clients, Vermillio can send take-down requests to sites such as YouTube or Instagram. Additionally, Vermillio says clients can monetize their data by licensing it.

People who sign up for the free version enter information about themselves such as their name, date of birth and social media handles on sites including Instagram or YouTube.

Then, Vermillio will use that information to build a “likeness model” to scour the Internet for potential red flags involving the user’s identity. Then Vermillio alerts the user to what exists online. For example, if someone has created a fake Instagram account of that user, Vermillio would flag that.

Users are notified of this type of content and can decide for themselves what they would like to allow, or take action to remove. If the user wants Vermillio to request take-downs of the inappropriate content, users would need to upgrade to a paid account, which starts at $10 a month and includes five monthly take down requests.

While many social media platforms give an option to users to flag problematic content, Vermillio said it is faster and more effective than having users go directly to YouTube or Instagram to rectify the situation. It has built a network of partners and can escalate take-downs in as quickly as an hour, the company said.

Vermillio executives said some real life examples of deep fakes include celebrity voices used to raise money for fake charities or terrorist organizations, and high school students creating fake pornography of their classmates.

“It’s affecting regular people in the sense that they’re getting scammed by deep fakes, but it’s also affecting teenagers, so people need to understand where they stand,” said Kathleen Grace, Vermillio’s chief strategy officer. “This is an easy way for them to do that.”

While fake social media profiles have existed for years, “generative AI just poured gasoline on it,” Grace said.

The company said hundreds of people use Vermillio’s services, but didn’t specify numbers. By the end of the year, the company expects to have thousands.

Neely said the company isn’t profitable and declined to share revenue figures. Time magazine reported that revenue from Vermillio’s TraceID has increased tenfold from April 2023 to April 2024. The company makes money through the paid versions of its service and licensing. Vermillio has raised $24 million in funding.

Hollywood companies and talent are navigating artificial intelligence in different ways.

Groups such as performers guild SAG-AFTRA are pushing for more state and federal protections against deepfakes. Some celebrities such as Academy Award-winning supporting actor Jamie Lee Curtis struggled to get a fake ad of her on Instagram taken down showing her falsely endorsing a dental product.

WME announced a partnership with Vermillio last year.

“The scale of the issue is extraordinary, so if you’re a rights holder, just trying to understand how much of these AI outputs are based on or utilized my data, my IP in some way, shape or form, is a massive need,” said Chris Jacquemin, WME’s head of digital strategy.

“They’ve obviously proven that TraceID can protect the most important, most high profile public figures in the world,” Jacquemin added. “Opening it up in a much broader application, I think is a huge step forward in really democratizing how anybody can start to police use of their likeness with respect to AI and AI platforms.”