Publication of intimate images containing “deepfakes” created by AI without the consent of legal offenders.
US President Donald Trump has signed a bill banning so-called “venge porn,” including images created using artificial intelligence.
The Take It Down Act, signed Monday, would make it a federal crime to publish intimate images of people without their consent, requiring social media platforms to remove such images within 48 hours when requested by victims.
The law, which passed the US Congress with almost indifferent support earlier this year, applies to realistic AI-created images and real photos known as “deepfakes.”
“The rise of AI image generation has led countless women to be harassed by deepfakes and other explicit images against their will. This is… wrong… very badly wrong,” Trump said at a signing ceremony at the White House in Washington, DC.
“It’s a very abusive situation…and today we make it completely illegal.”
First Lady Melania Trump, who supported the law after her husband returned to the White House, described the law as “a powerful step in our efforts to ensure that all Americans, especially young people, can feel better protected from abused images and identities.”
While enjoying the rare, bipartisan support and the support of many organisations dedicated to combating sexual harassment and assault, the law has elicited criticism from digital rights groups on the basis of privacy and anti-sensorship.
The Electronic Frontier Foundation said the provisions in the law to remove flagged materials from the Internet “pose a risk to free expression, user privacy and due process without addressing the issues it claims to be resolved.”
“Legal content, including satire, journalism and political speech, can be miscensored,” the group said in a February statement.
“In the strict legal time frame, apps and websites must remove content within 48 hours. This means that online service providers, especially small providers, must comply very quickly to avoid legal risks that cannot verify claims. Instead, automated filters are used to catch duplicates, but these systems are notorious for flagging legal content from fair statements.”