Photo credit: www.cbsnews.com
Meta has recently taken action to eliminate numerous deceptive, sexualized images of well-known female celebrities and athletes following an investigation by CBS News that revealed a disturbing prevalence of AI-manipulated deepfake content on Facebook.
Images of notable figures such as Miranda Cosgrove, Jeanette McCurdy, Ariana Grande, Scarlett Johansson, and former tennis champion Maria Sharapova have surfaced across multiple Facebook accounts, many of which received extensive engagement, including hundreds of thousands of likes and shares.
In a statement to CBS News, Meta spokesperson Erin Logan confirmed, “We’ve removed these images for violating our policies and will continue monitoring for other violating posts. This is an industry-wide challenge, and we’re continually working to enhance our detection and enforcement technology.”
Reality Defender, a platform specializing in identifying AI-generated content, conducted an analysis of some problematic images and found that most were deepfakes. These involved the use of AI to generate sexualized, clothing-altered versions of these celebrities while retaining the real backgrounds from legitimate photographs. A few of the images appeared to stem from image-stitching techniques that do not incorporate AI.
Ben Colman, the co-founder and CEO of Reality Defender, commented, “Almost all deepfake pornography lacks the consent of the individuals depicted. The proliferation of such content is escalating rapidly, especially since measures designed to combat it are infrequently enforced.”
CBS News has reached out to the aforementioned actors for their reactions to the issue, with Johansson’s representative stating that she would not be commenting.
Meta’s Policies and Current Challenges
Under its Bullying and Harassment policy, Meta prohibits “derogatory sexualized photoshop or drawings” across its platforms. The guidelines also outlaw adult nudity, sexual activity, and any form of adult sexual exploitation, aiming to curtail the distribution or threats of sharing non-consensual intimate imagery. Moreover, the company has adopted “AI info” labels to identify content manipulated using AI.
Nevertheless, the efficiency of Meta’s approach to regulating this type of content is under scrutiny. CBS News discovered many AI-generated, sexualized images of Cosgrove and McCurdy remaining publicly accessible, despite being reported as violations of Meta’s policies.
One such deepfake image of Cosgrove remained visible over the weekend, promoted by an account boasting 2.8 million followers.
Both Cosgrove and McCurdy, actresses who rose to fame as child stars on the Nickelodeon series iCarly, are among the most frequently targeted celebrities for deepfake imagery, as observed in CBS News’ analysis of social media content.
Calls for Stronger Regulations
The Oversight Board of Meta, which includes experts in human rights and free speech, expressed concerns regarding the inadequacy of the company’s existing regulations related to sexualized deepfake content. They highlighted recommendations made over the past year, suggesting that Meta update its guidelines to explicitly include terms like “non-consensual” and broaden the scope to encompass various manipulation techniques, including AI-generated content.
Additionally, the board proposed merging the rules surrounding “derogatory sexualized photoshop” with the company’s guidelines on Adult Sexual Exploitation to enhance enforcement mechanisms
.
When asked about the Oversight Board’s recommendations, Meta referenced its transparency website, indicating they are evaluating three out of four suggestions provided by the board. However, the company has decided against revising the “derogatory sexualized photoshop” policy to include “non-consensual.” According to their statement, moving these regulations under Adult Sexual Exploitation guidelines is also deemed unlikely at this time.
Meta acknowledged ongoing considerations regarding how to effectively denote a lack of consent in AI-generated imagery, as well as potential reforms to their Adult Sexual Exploitation policies that would align more closely with the board’s suggestions.
Michael McConnell, co-chair of the Oversight Board, stated, “Non-consensual deepfake intimate images represent a significant violation of privacy and personal dignity, disproportionately impacting women and girls. These images constitute not merely the misuse of technology but rather a form of abuse that can carry long-term ramifications.”
The Oversight Board continues to monitor Meta’s actions and is advocating for more robust protections, expedited enforcement, and increased accountability from the platform.
Meta is not alone in grappling with the challenge of rampant sexualized deepfake content on social media. Last year, the platform X, led by Elon Musk, temporarily suspended searches related to Taylor Swift after AI-generated explicit images imitating the singer circulated widely, garnering millions of views.
The safety team of platform X reiterated their commitment, stating, “Posting Non-Consensual Nudity (NCN) images is strictly prohibited and we maintain a zero-tolerance policy towards such content.”
A recent study from the U.K. government revealed a swift increase in the prevalence of deepfake images across social media channels, with predictions estimating that eight million deepfakes could be shared this year—up from just 500,000 in 2023.
Source
www.cbsnews.com