In an unprecedented legal move, a court in the United Kingdom has banned a convicted sex offender from using or accessing artificial intelligence tools capable of generating images, video, audio or text. The ruling came as part of a sexual harm prevention order issued in the case of Anthony Dover, 48, who was found guilty of creating over 1,000 indecent images of children.
Legal and tech experts are calling this a landmark case, as it marks the first known instance in the UK of a court specifically prohibiting the use of AI-powered tools, such as Stable Diffusion, by a criminal offender. The decision shines a spotlight on the growing concern over how generative AI systems could be misused to create illegal content like child sexual abuse material (CSAM).
“This case should sound the alarm that criminals producing AI-generated child sexual abuse images are like one-man factories, capable of churning out some of the most appalling imagery,” said a spokesperson for the Internet Watch Foundation (IWF), a UK-based charity that works to minimize the availability of online child sexual abuse content. The IWF noted it has seen a “slow but continual” rise in the proportion of AI-generated CSAM, though official statistics are not yet available.
While it remains unclear if Dover actually used AI tools to generate the illegal images he was convicted of creating, the court order specifically named Stable Diffusion as falling under the ban. Stable Diffusion, an AI image generation system, has previously been cited in UK court cases involving the creation of CSAM.
In response, a representative from Stability AI, the company behind Stable Diffusion, stated they prohibit the use of their software for unlawful activities like producing child abuse images. They added that the version of Stable Diffusion referenced in prior legal cases predates the company securing an exclusive license for the tool in 2022. Stability AI said the concerning images were likely made using the earlier open-source version.
Under the terms of the sexual harm prevention order, Dover can only use generative AI tools if he obtains advance permission from law enforcement for specific named tools. In addition to the AI restrictions, he was sentenced to a community order and fined £200 ($247).
Putting Limits on Lawbreakers' Tech Use
Imposing constraints on sex offenders' internet and technology usage is not a new practice for UK courts. Judges have previously banned criminals from using certain messaging apps or browsing the web in “incognito” mode in an effort to stymie further illegal activities and protect public safety.
However, the addition of restrictions on AI tools opens up new logistical and ethical questions as this emerging technology becomes increasingly integrated into everyday life and work. It's conceivable that those subject to such bans could be excluded from jobs that involve the use of generative AI systems that are becoming more commonplace in certain industries and roles.
While few may have sympathy for criminals facing these consequences, digital rights advocates argue there need to be clear guidelines and standards around the use of AI restrictions in sentencing. Lack of specificity in court orders could lead to confusion for offenders trying to comply and challenges for probation officers tasked with enforcement. There are also concerns it could hamper rehabilitation efforts and reintegration into society if bans are overly broad.
A Turning Point for AI Regulation?
Dover's case comes at a pivotal moment in the public and policy discourse around the explosive rise of generative AI tools. The relative ease with which these systems can now be used to create fake pornographic videos and images, often without the consent of those depicted, has sparked calls for tighter regulation.
While the creation of CSAM is unambiguously illegal in the UK and most countries, the laws governing the generation of explicit deepfake content involving adults have been murkier – until now. The UK government recently announced that producing pornographic deepfakes without the consent of the individuals represented will be a criminal offense if done with the intent to cause “alarm, humiliation or distress.” Sharing such nonconsensual material was already illegal.
Read more about the announcement Here
Other European governments are also grappling with how to crack down on malicious deepfakes as complaints arise. In April, the Prime Minister of Italy, Giorgia Meloni, filed a defamation lawsuit against two men allegedly responsible for creating and spreading pornographic videos that featured her digitally altered likeness.
As disturbing examples of nonconsensual explicit deepfakes targeting high-profile figures make headlines, some legal experts believe Dover's case could prove to be a bellwether – potentially ushering in more proactive court orders restricting AI use by offenders. A Crown Prosecution Services spokesperson said they will ask courts to “impose conditions, which may involve prohibiting use of certain technology” when they perceive an ongoing threat to public safety, especially children.
Digital rights groups generally agree on the need to combat egregious abuses of generative AI tools. However, some caution against an over-reliance on outright bans, advocating instead for more nuanced approaches focused on educating users about ethical boundaries, improving AI systems' ability to detect CSAM, and supporting the development of “radioactive” data sets that would taint any model trained on them.
As lawmakers worldwide race to craft policies to rein in the misuse of generative AI, test cases like Dover's will be closely watched. How the courts balance public protection with fairness for defendants, and keeping pace with technological change, could have far-reaching implications for the future of AI governance. One thing is certain: the genie is out of the bottle, and the disruptive impact of AI-generated content, both positive and negative, will only continue to accelerate. Policymakers face the daunting task of building guardrails for a technology that is evolving faster than the law.
