In a disturbing trend, online trolls on the infamous message board 4chan have been utilizing artificial intelligence (A.I.) tools to spread racist and offensive content about individuals who have appeared before the Louisiana parole board. These trolls have been using A.I.-powered image generators and audio editors to manipulate screenshots and create fake images and audio, including making a doctor who testified during a parole hearing appear naked.

Daniel Siegel, a graduate student at Columbia University who studies the exploitation of A.I., has been documenting this activity on 4chan for several months. While the manipulated images and audio have not spread widely beyond the confines of 4chan, experts who monitor fringe message boards warn that this could be a glimpse into how malicious internet users may utilize sophisticated A.I. tools to amplify online harassment and hate campaigns in the future.

Callum Hood, the head of research at the Center for Countering Digital Hate, explains that fringe sites like 4chan often serve as early indicators of how new technology will be used to project extreme ideas. These platforms attract young individuals who are quick to adopt new technologies like A.I. to spread their ideologies into mainstream spaces. Unfortunately, these tactics are also adopted by some users on more popular online platforms.

One of the problems resulting from A.I. tools that experts have discovered on 4chan is the creation of artificial images and A.I. pornography. Tools like Dall-E and Midjourney can generate images from simple text descriptions, but a new wave of A.I. image generators specifically focus on creating fake pornography, including removing clothes from existing images. This poses a significant challenge for regulators and technology companies, as there is currently no federal law prohibiting the creation of fake images of people. The Louisiana parole board has initiated an investigation in response to Siegel's findings on 4chan, but they are limited in their actions by the absence of clear legal guidelines.

To combat this issue, some states have taken steps to address A.I.-generated pornography without consent. Illinois expanded its revenge pornography law to allow targets of nonconsensual pornography made by A.I. systems to sue creators or distributors. California, Virginia, and New York have also passed laws banning the distribution or creation of A.I.-generated pornography without consent.

Another problem arising from A.I. tools is the cloning of voices. A tool released by A.I. company ElevenLabs last year allowed for the creation of convincing digital replicas of someone's voice. Almost immediately after its release, 4chan users began circulating clips of a fake Emma Watson reading Adolf Hitler's manifesto. Similarly, using content from the Louisiana parole board hearings, 4chan users have shared fake clips of judges uttering offensive and racist comments about defendants. Many of these clips were generated using ElevenLabs' tool. While ElevenLabs implemented limits on the tool, such as requiring users to pay for access, the spread of A.I.-created voices has not been curtailed significantly. Social media platforms like TikTok and YouTube have now implemented labels on some A.I.-generated content, and President Biden issued an executive order calling for the labeling of such content and the development of standards for watermarking and authenticating A.I.

The issue of open-source A.I. tools has also come to the forefront. Meta, formerly known as Facebook, has been releasing its software code to researchers as part of its strategy to gain a foothold in the A.I. race. However, when Meta released its large language model, Llama, to select researchers, the code quickly leaked onto 4chan. Users on the platform modified the code to remove guardrails, resulting in the creation of chatbots capable of producing antisemitic ideas. This incident highlights the potential dangers of open-source A.I. tools in the hands of technologically savvy individuals. Meta defends its release strategy, stating that it allows for a balance between responsibility and openness.

In recent months, language models developed from open-source A.I. tools have been used to echo far-right talking points and create sexually explicit content. This further emphasizes the need for regulation and responsible usage of A.I. tools to prevent the spread of harmful and offensive content.

As the exploitation of A.I. for malicious purposes continues to evolve, it is crucial for regulators and technology companies to stay vigilant and implement measures to safeguard against the misuse of these powerful tools.