AI Pulse by Inblix

OpenAI busts UAE-linked firm using ChatGPT to fake support for Ghana's VP

OpenAI Blog · Feb 1, 2025 · 2 min read · Read original article →

Curated by the Inblix editorial team


Featured image for article: OpenAI busts UAE-linked firm using ChatGPT to fake support for Ghana's VP

OpenAI has banned a cluster of accounts linked to a commercial entity in the UAE and Ghana for a clumsy covert influence operation targeting Ghana’s 2024 presidential election. The operation, centered on a website called Empowering Ghana, used ChatGPT to churn out pro-Vice President Mahumudu Bawumia content and fake engagement across multiple platforms.

The company behind the scheme, DigitSol, didn’t exactly cover its tracks well. The Empowering Ghana website, which claimed to be a youth initiative, listed an Australian street address, an apparently bogus phone number, and a link to a WordPress plugin provider. It’s the kind of sloppy tradecraft that makes you wonder if they were even trying to hide. OpenAI’s report details how the group used its models to not only write long-form articles but also to generate batches of 30 short comments praising Bawumia’s policies—all posted by single social media accounts, some with no profile pictures and others with images stolen from across the web.

The engagement was almost entirely fake. The Empowering Ghana Instagram account boasted around 900,000 followers, but Instagram’s own transparency tools showed 54% of them also followed a random account focused on Australian legislation. That’s roughly half a million accounts with a deeply improbable dual interest. As OpenAI dryly notes, the authentic engagement on most posts was likely between zero and two real comments, not the 30-32 the operation tried to simulate. The entire network went dark on December 8, 2024, the day after the election.

This isn’t a sophisticated state-backed espionage campaign. It’s a textbook example of what the industry calls “cheapfakes”—low-budget, low-impact manipulation. OpenAI rated the operation a 2 on its Breakout Scale, meaning activity existed on multiple platforms but there’s zero evidence real people picked it up or amplified it. The X account had just 1,535 followers, and YouTube videos pulled in double-digit views. No mainstream media touched the content. What’s genuinely notable here is the blatant commercial motive: this wasn’t a shadowy intelligence agency, but a business for hire, bringing the gig-economy ethos to political manipulation. It’s a reminder that the barrier to entry for running a textbook influence op is now just the cost of an API key and a stunning lack of scruples.

💡 Key Takeaways

  1. DigitSol, a UAE and Ghana-based commercial entity, was the actor behind the operation, not a state-backed group, highlighting how AI-powered disinformation is now a for-hire service.
  2. The operation's Instagram following of 900,000 was largely fraudulent, with over half of the followers sharing a bizarre link to a single Australian legislative account.
  3. OpenAI rated the operation's real-world impact as minimal (Category 2 out of 6), as it failed to gain any amplification from real users or mainstream media.

Keep reading: See related articles below for more coverage on this topic.

Get smarter about AI

The sharpest AI news, curated daily. Delivered free to your inbox.

Learn more

Glossary terms

← Back to all articles