Facebook is no newcomer to the task of regulating and addressing misinformation on its platform. For a considerable period, the company has utilized machine learning and artificial intelligence systems to complement its human-led moderation endeavors. At the outset of October, Facebook expanded its machine learning capabilities into its advertising domain by introducing a set of experimental generative AI tools. These tools are capable of tasks such as generating backgrounds, adjusting images, and creating captions for video content used in advertising.
As per a report from Reuters, Meta, the parent company of Facebook, has made a specific decision not to provide these generative AI tools to political advertisers in anticipation of a contentious national election cycle. Although this decision aligns with much of the social media landscape, Meta has not yet publicly updated its advertising standards to reflect this choice.
It’s worth noting that Meta does permit exceptions to this rule. The ban on these tools is primarily focused on preventing “misleading AI-generated video content in all contexts, including organic non-paid posts,” with a provision for parody or satire content, according to Reuters. Currently, these exceptions are being reviewed by Meta’s independent Oversight Board, particularly in a case involving a disputed video of President Biden that Meta argued was not generated by AI.
In July, Facebook, along with other prominent Silicon Valley AI companies, entered into voluntary commitments outlined by the White House, which entail implementing technical and policy safeguards in the development of future generative AI systems. These commitments include expanding adversarial machine learning efforts to identify and address problematic model behavior, sharing trust and safety information within the industry and with the government, and establishing a digital watermarking system to authenticate official content and clarify its non-AI origin.
