US senators demand answers from tech giants over explicit deepfakes
What's the story
A group of US senators has written to tech giants Meta, Alphabet, X, Snap, Reddit, and TikTok. The letter asks these companies to prove that they have "robust protections and policies" in place against sexualized deepfakes on their platforms. The senators also want the firms to detail how they plan to tackle this growing problem.
Document preservation
Senators ask for preservation of documents
The senators have also asked the companies to preserve all documents and information related to the creation, detection, moderation, and monetization of sexualized AI-generated images. This includes any policies they may have in place on these matters. The request coincided with X announcing an update for its AI tool Grok, which now prevents it from editing real people in revealing clothing.
Guardrail effectiveness
Concerns over platforms' guardrails against non-consensual imagery
The senators have expressed concern over the effectiveness of platforms' guardrails to prevent users from posting non-consensual, sexualized imagery. They noted that despite many companies having policies against such content and AI systems claiming to block explicit pornography, users are still finding ways around these protections. The letter highlighted this issue with specific reference to Grok's ability to generate sexualized and nude images of women and children.
Platform challenges
Deepfakes' history and current status on various platforms
Deepfakes first gained notoriety on Reddit, where a page featuring synthetic porn videos of celebrities went viral before being taken down in 2018. Now, sexualized deepfakes targeting celebrities and politicians have multiplied on TikTok and YouTube. Meta's Oversight Board flagged two cases of explicit AI images of female public figures last year.
Information request
Senators's letter seeks detailed information from tech companies
The senators' letter asks the companies for a range of information. This includes their definitions of "deepfake" content and "non-consensual intimate imagery," as well as how they enforce policies against non-consensual AI deepfakes. They also want to know what filters or guardrails have been put in place to stop the creation and distribution of deepfakes, and how these companies prevent users from profiting off such content.
Company responses
Tech companies respond to senators' letter
In response to the senators' letter, X pointed to its announcement of an update for Grok. A Reddit spokesperson said, "We do not and will not allow any non-consensual intimate media (NCIM) on Reddit." They added that Reddit strictly prohibits NCIM, including depictions that have been faked or AI-generated. However, Alphabet, Snap, TikTok, and Meta have yet to respond on this matter.