Meta's AI Fails to Block Deepfake Scam Ads Targeting Polish Businessman
Meta's content moderation systems on Facebook and Instagram failed to block deepfake scam advertisements that used the likeness of Polish InPost president Rafał Brzoska and his wife to defraud users. Despite court orders and decisions by Poland's data protection authority requiring removal, the fraudulent ads continued circulating and even reached Brzoska's children, prompting his public criticism and the Polish government's intervention.
What happened
Rafał Brzoska, president of InPost, reported that Meta's Facebook and Instagram platforms displayed illegal deepfake advertisements and scam content that unlawfully used his and his wife's likenesses to defraud users.
Despite court orders and decisions by the Polish Data Protection Office (UODO) mandating removal of this content, the false advertisements continued to be published and reached Brzoska's children. Meta's algorithms and human moderators failed to detect the fraudulent content after initial reports.
Reported impact
The scam ads targeting Brzoska's identity were used to extort money from the public. The content reached Brzoska's own children despite legal orders for its removal.
In response to Brzoska's public criticism of Meta and disclosure of Polish managers' names, Meta temporarily suspended his account while denying responsibility for the problem. The Polish government subsequently demanded explanations from Meta.
Index analysis
According to Panoptykon Foundation expert Katarzyna Szymielewicz, Meta's responsibility under Polish and European law is limited—it only applies once the platform is notified of illegal content. Meta can argue its algorithms did not detect the deepfakes and moderators lacked sufficient context to identify the scam.
Under the EU's Digital Services Act (DSA), Meta should design its algorithms and interfaces to prevent the spread of disinformation, scams, and illegal content. Szymielewicz assessed that Meta does not do this effectively, as demonstrated by the Brzoska case.
Governance controls
Meta's content moderation relies on algorithmic detection and human review. The case reveals that Meta's systems failed to automatically identify and remove deepfake advertisements even after court orders and UODO decisions required their removal.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.