When a video of former prime minister Raila Odinga endorsing a cryptocurrency surfaced on his official X account last Thursday, it stunned the country.
With about 4.9 million followers, Odinga's verified account lent the clip instant credibility, and the footage, which appeared to show him championing a so-called 'Kenya Digital Token', spread rapidly across multiple social media networks.
"Kenya is stepping up to lead Africa into the crypto revolution, embracing digital finance and shaping a more crypto-friendly future," the video proclaimed in Odinga's likeness and voice.
But it was all fake. Within hours, it emerged that Odinga’s account had been hacked and the video was a deepfake, crafted to promote a dubious digital coin.
The video was eventually deleted, but not before exposing how even Kenya’s most prominent figures can be turned into unwitting vehicles of fraud in the age of artificial intelligence (AI).
This was, however, not the country's first brush with synthetic deception. In late February this year, Foreign Affairs Principal Secretary Korir Sing’oei was forced into an embarrassing apology after he shared on his own X account a fabricated CNN-style broadcast.
The AI-generated clip falsely portrayed Kenya’s role in Sudan’s peace talks and was swiftly debunked, but the embarrassment lingered.
At the time, many dismissed it as an isolated misstep, but the Odinga hack brings to the fore yet another layer of proof that deepfakes are increasingly emerging as an active tool for fraud and disinformation.
“Deepfakes can destabilise democracies by eroding trust in leaders, institutions and even the media itself,” warns tech policy analyst Kennedy Kamande.
“If a video of Odinga can circulate unchecked, imagine the consequences if such content is dropped on the eve of an election. It could be catastrophic.”
While analysts concur that it could oftentimes be challenging to tell fake from real, especially due to the enhanced sophistication of modern-day tech tools, there are usually visible indications to check out for before distributing content.
“There are some practical clues to look out for, such as watching for unnatural facial movements, off lip-syncing, lighting inconsistencies, and using reverse video search and fact-checking tools. You can also check for traces of the same content online,” says social media analyst Egline Samoei.
Kenya’s legal framework offers little comfort. While the Computer Misuse and Cybercrimes Act touches on misinformation, it does not explicitly address AI-generated manipulation.
In a number of countries, some governments are experimenting with watermarking AI-generated content, but experts warn that watermarks can be stripped out.
While tendering his apology in February, PS Sing’oei had announced the government had been working with partners and tech firms to explore the potential of watermarking AI-generated content as part of a bid to ensure that such material does not pose as human-generated.
“Additionally, we are designing – with the collaboration of partners and big-tech companies – the School on AI Diplomac at the Foreign Services Academy to equip policymakers better to navigate these emerging challenges,” said the PS at the time.
But according to industry analysts, Kenya needs a layered response that combines regulation, tech partnerships, and aggressive digital literacy campaigns.
“The watermarks could certainly help in identifying AI-generated content, but it’s not a complete solution because those watermarks can be erased or cropped out,” says Ms Samoei.
On his part, Kamande calls for an update of the country’s legal apparatus, in addition to upscaling media literacy amongst the population.
“The country must update its legal frameworks to criminalise deepfake abuse. Offenders should face strict penalties, and law enforcement agencies must be equipped with digital forensic tools to track and prosecute perpetrators,” he notes.
