At 9:03 a.m. on July 9, 2026, a Delta Air Lines flight lifted off from Accra’s international airport carrying one of Ghana’s most-followed social media personalities. Destination: New York, and a US federal courtroom. Frederick Kumi — better known online as Abu Trica, with over 100,000 followers — now faces charges that sound almost cinematic: conspiracy to commit wire fraud and money laundering, tied to more than $8 million allegedly stolen from elderly Americans.
What sets this case apart from the usual West African romance-scam headlines isn’t the sum. It’s the tool: artificial intelligence.
The man behind the name
Kumi, 31, is from Swedru, a town in central Ghana. According to an indictment from the US Attorney’s Office for the Northern District of Ohio, he — also known as Emmanuel Kojo Baah Obeng — was part of a criminal network that had been targeting elderly American victims in romance scams since 2023.
He was arrested in Ghana on December 11, 2025, and spent months fighting extradition. He lost that fight, though not without a chaotic final chapter.
When AI does the flirting
Prosecutors allege the network used AI software to build convincing fake identities and cultivate deeply personalized emotional relationships with victims. Forget the clumsy, typo-riddled scam messages of a decade ago — this is generated photos, emotionally-tuned conversation, and a scaled-up, almost industrial approach to manufacturing intimacy.
The playbook still ends the same way: months of trust-building, then a sudden financial emergency — a medical bill, a blocked visa, a customs fee. What’s changed is that AI now makes every message feel a little more natural, a little more real.
A messy extradition
The court ruling wasn’t the end of the drama. On July 2, 2026, Accra’s High Court approved the US extradition request — but the way it was carried out sparked outrage. Kumi’s lawyer, Oliver Barker-Vormawor, says his client was put on a plane while an emergency motion to halt the extradition was still pending in court, and that he was denied access to his client for days, unaware of where he was even being held.
His legal team is now framing the case as a constitutional matter, accusing the Ghanaian government of sidestepping judicial oversight — turning a fraud case into a national debate about sovereignty under US pressure.
Part of a bigger pattern
Abu Trica isn’t operating alone in the bigger picture. US authorities recently extradited several other Ghanaian nationals tied to a separate network accused of stealing more than $100 million through romance scams and business email compromise schemes. The pattern repeats: isolated, often older victims, targeted with near-surgical precision, and large sums funneled through sophisticated laundering chains.
Why this matters beyond America
It’s tempting to file this under “not my problem” — American victims, American courts. But these networks increasingly target dating platforms used by the African diaspora and singles across the continent too. The warning signs haven’t changed much since before AI entered the picture:
- A profile that looks too perfect — model looks, a doctor or engineer “on a humanitarian mission”
- An “I love you” within the first two weeks
- Consistent refusal to do a live video call
- A sudden financial emergency (customs, accident, visa, hospital bill)
- Pressure to move the conversation off-platform to WhatsApp or Telegram
The difference now is that the photos can be entirely synthetic, and the conversation is unnervingly natural. The good news: a live video call is still, for now, close to foolproof. No AI can yet sustain a spontaneous, fluid real-time video conversation.
This case unsettles me for reasons that go beyond the dollar figure. What bothers me is how fast a technology meant to bring us closer — being able to talk to anyone, anywhere, anytime — has become an industrialized manipulation tool. We tend to quietly blame victims in these stories (“she should have known better”), but the shame needs to shift. Victims aren’t naive — they’re human, and AI is specifically built to exploit that.
Do you think is the responsibility entirely on us as users?