The first lady has platformed an important piece of legislation to keep the AI horror-scape at bay.
The first lady has platformed an important piece of legislation to keep the AI horror-scape at bay.
F irst Lady Melania Trump recently held a roundtable in support of legislation that’s crucial to prevent the creation and dissemination of “revenge porn.”
If passed, the Tools to Address Known Exploitation by Immobilizing Technological Deepfakes on Websites and Networks Act, a.k.a., the Take It Down Act, would
protect and empower victims of non-consensual intimate image abuse — also known as “revenge pornography.” The bill would criminalize the publication of non-consensual intimate imagery (NCII), including AI-generated NCII (or “deepfake pornography”), and require social media and similar websites to have in place procedures to remove such content upon notification from a victim.
The bipartisan bill was introduced by Senator Ted Cruz (R., Texas) last year and is co-sponsored by Senator Amy Klobuchar (D., Minn.) along with several other senators from both parties. The act might already be law had not Senator Cory Booker (D., N.J.) slowed its progress — he filed a last-minute objection last fall.
We are on the edge of a revenge-porn epidemic. Melania Trump is right to say that the Take It Down Act deserves congressional priority. Democrats need to act now, in cooperation with Republicans, to protect women and girls across the country.
Revenge porn may seem like a niche issue, but the phenomenon is growing at an exponential rate — especially in schools. In the early 2000s, during the first wave of distress over revenge porn, the intimate photos had to be taken in real life before they could be leaked online. Now, through advances in AI, anyone with access to the internet can turn an innocent photo into a non-consensual intimate image (NCII).
An app called DeepNude, which uses AI to “denude” any photo of a clothed person, was one of the first that allowed users to make NCIIs at the touch of a button. Released in 2019, the app was quickly shut down after the team that created it acknowledged that “the probability that people will misuse it is too high.”
But the cat was out of the bag. Hundreds of copycat apps with the same function have since popped up online. And, more recently, hundreds of apps have emerged that enable the creation of entirely synthetic images or videos — “deepfakes” — modeled after the visage of a particular person.
Deepfakes are created when an AI program is “trained” on a collection of photos or videos of a reference subject — or, in the case of NCIIs, a reference victim. Once the AI learns what the reference individual looks like, it can swap that person’s face, voice, hands, etc. with those in target photos or videos. Advanced programs don’t even need target content — a text prompt is all it takes to create a purely synthetic image or video that portrays the reference subject in an imagined context.
While watchdog groups, media outlets, and state legislatures have responded to the potential of deepfake technology to spread disinformation and swing elections, the dangers of deepfake NCIIs have been largely neglected.
Of course, all should be wary of political deepfakes which can be used to keep voters from the ballot boxes or smear candidates on the cusp of an election. But a deepfake of a governor or representative proclaiming, say, his allegiance to the KKK is easier to debunk than a deepfake of a teenage girl in an NCII. Regardless, even if the images are taken down, the damage has already been done.
As of 2024, 15 percent of students in K–12 schools say they know of an NCII depicting a peer in their school. To the surprise of none, this problem poses a grave danger for young women especially. A software company that detects deepfakes, Sensity AI, tracked online deepfake videos and found that 90 to 95 percent of them were nonconsensual porn, and 90 percent of those videos were of women.
At the Take It Down roundtable, two girls, Francesca Mani and Elliston Berry, shared their own stories of being victims of AI-created NCIIs.
Berry — who is 15 years old — spoke of how a classmate created AI nudes of her from an innocent picture on Instagram:
On October 2, 2023, I woke up to messages from a friend notifying me that these photos of me were circulating social media. Pictures from a past Instagram photo with a nude body and my face attached made from AI. Fear, shock, and disgust were just some of the many emotions I felt. I felt responsible, and began to blame myself, and was ashamed to tell my parents, despite doing nothing wrong.
Berry and her parents tried to get the photos taken down from Snapchat, where they were hosted, but received no help. “It wasn’t until Ted Cruz’s office heard my story and reached out to Snapchat personally, that my photos were removed,” Berry said. “We need to hold Big Tech accountable.”
Mani, who had a similar experience, emphasized that there needs to be a way to hold perpetrators accountable in the justice system and in schools across the country.
The Take It Down Act would make the publication of an NCII a federal crime and implement harsher penalties against the perpetrator if the victim is a minor. The law would also require all platforms for user-generated content to have a process of notice and removal for any NCIIs that are published on their site.
To date, there is no comprehensive federal law that criminalizes the creation or dissemination of deepfake pornography. AI legal expert Amy Swaner notes:
Revenge porn laws, originally designed to criminalize the nonconsensual distribution of real intimate images, struggle to address the unique challenges posed by AI-generated deepfake pornography. Many existing laws require that the images be authentic, meaning they must depict the actual victim rather than a synthetic creation. This loophole allows perpetrators to claim that no real photo or video was used, evading liability. [Emphasis in original.]
Because deepfakes can be entirely synthetic, although made to look like a particular individual, users who create and publish deepfake NCIIs can evade legal penalties in most states. In 2019, Virginia became the first state to amend its revenge porn laws and incorporate provisions against NCIIs. Texas and California have followed suit. However, most women and girls across America remain unprotected from NCIIs.
Several states lack penalties for the publication of deepfake NCIIs depicting children. So far, 13 states have not enacted laws against AI-generated child sexual abuse material (CSAM). Currently, there is no federal law prohibiting AI-generated CSAM.
The Take It Down Act would protect women and girls in all 50 states from the grave harms of NCIIs and offer recourse to victims. It cannot be passed soon enough.