The Republic of Ghana lodged a $7.5 billion civil suit in the US District Court for the Northern District of California on Tuesday, alleging that Aperture Intelligence built and sold an algorithm trained on millions of images and personal records taken from Ghanaian social media and public registers without consent. The suit names the company and three of its subsidiary services and asks for damages, a binding data‑protection compliance order and a block on the use of Ghanaian biometric features in future models.
Ghana’s claim, filed by the attorney‑general’s office and three human‑rights groups, says the model was used by local financial firms, private security contractors and two municipal police units, and that errors in the system led to wrongful detentions, loan denials and lost work for thousands. Lawyers for the plaintiffs say the action is intended both to recover compensation and to force structural changes in how foreign firms use Ghanaian data.
The suit’s central allegations
According to the 234‑page complaint, a forensic review by Accra Data Lab — commissioned by the attorney‑general — found that Aperture’s training set included roughly 120 million images scraped from Ghanaian social networks, online news comments and an archival leak of municipal voter lists. The filing alleges the company sold a ‘‘citizen‑risk’’ product globally that assigned negative scores to Ghanaian users at rates three times higher than for other nationalities, and that automated decisions based on those scores compounded poverty and restricted mobility.
Marija Vincas, lead counsel for the Ghanaian plaintiffs, said the government had exhausted diplomatic channels before suing. "We sought redress directly for people who lost livelihoods because a black‑box system treated them like security risks," Vincas said. "This is about accountability where voluntary fixes failed."
How researchers tied models to harm
Researchers working with the attorney‑general compared Aperture outputs against independent ID registries and found thousands of false positives in identity matches, the complaint says. Joris A. Hofman, a data scientist at Accra Data Lab who analysed the models, said the firm’s face embeddings correlated with metadata such as place of birth and economic indicators more strongly in Ghana than in five comparison countries. "The algorithm was not neutral — it amplified historical administrative biases present in the records used for training," Hofman said.
We found families denied loans because an automated score flagged a child’s school photo as a fraud risk months later.
Marija Vincas, lead counsel
Aperture Intelligence issued a brief statement through a spokesperson saying it would defend the case vigorously and that its products comply with applicable law. Lilla Adams, identified in the filing as a company communications lead, said the firm was ‘‘committed to improving systems’’ and disputed authorship of some of the analytic claims in Ghana’s filing. The company declined to comment on operational contracts alleged in the suit.
Wider consequences for African tech policy
Legal experts in Accra and abroad said the case could force multinational AI developers to negotiate directly with states over population‑scale datasets. Asirwanda Hutasoit, a policy analyst with the Pan‑African Privacy Forum, said regulators in at least three West African countries were already reviewing contracts with foreign model providers. "If Ghana wins enforceable remedies, firms will have to adopt new licensing terms or move data‑intensive work out of high‑risk jurisdictions," Hutasoit said.
The filing is scheduled for an initial hearing in December. For now, Ghana’s authorities say they will continue parallel domestic investigations into how municipal agencies purchased and deployed automated scoring tools. The case marks what legal scholars called a turning point: an African state using the US legal system to ask a global technology seller to answer for local harms.
