Following the Taylor Swift deepfake incident, Google Bard’s new AI image generation feature sparked concerns about content safety. Despite Google’s claim of implementing filters to prevent generating images of named individuals, a user, Russ Silberman, successfully created an image resembling Swift. His prompt subtly referenced her, including specific details like a football jersey and a hot dog, testing Bard’s supposed guardrails. This incident highlights vulnerabilities in AI moderation as Silberman noted he could circumvent these protections by repeatedly clicking “regenerate draft.” Google acknowledged that while they strive to limit harmful content, their systems aren’t infallible. This situation underscores the potential risks linked to AI-generated images, which can easily be manipulated, raising alarm about privacy and representation in digital media. As deepfakes proliferate, the challenges of legal implications and ineffective platform safeguards become increasingly urgent, emphasizing the need for improved technologies and policies in AI usage.
Source link
Share
Read more