Band 18, Heft 2 - Tort Liability for Deepfakes: A Comparative Perspective

Generative AI technology is taking the world by storm. The development of technology enabling creation and manipulation of content has give facilitated a substantial rise in the proliferation of deepfakes. Whereas in the past content creation and manipulation required a certain level of expertise, today deepfake technology is easily accessible and enables the quick and seamless creation of highly believable content. This technology has many positive applications: for example, in healthcare, education, cultural preservation and the entertainment industry. But deepfake technology is also used to deceive and cause harm. Deepfake technology is used to create sexual deepfakes that humiliate and harm primarily women and girls, to engage in fraudulent activities, and to generate disinformation undermining trust in democratic processes and institutions. This Article suggests analyzing the deepfake dilemma through the lens of the doctrine of unjust enrichment. Under this doctrine a party which has become enriched at the expense of another must make restitution of benefits it received. An enrichment-based approach may offer several advantages, particularly when compared to harm-based remedies. First, it may be easier to identify the defendant in an unjust enrichment case, as the defendant is the company developing the underlying technology and there is no need to identify the individual who created the deepfake. Second, a lawsuit under unjust enrichment may be filed by various plaintiffs, thus, obivating the need to identify the individual harmed by a deepfake. Finally, it may be easier to identify and quantify the profits which are monetary, current and concentrated within a small number of companies. We argue that under certain circumstances viewing the deepfake challenge through the lens of unjust enrichment can allow for a realignment of the financial incentives of the companies developing deepfakes technology with broader social goals and values, encouraging them to develop technology that is less harmful and more responsible.

As deepfake technology becomes increasingly sophisticated and accessible, American lawmakers are responding with a flurry of urgent legislative action to address its potential harms. Our 50-state survey of proposed and enacted deepfake legislation reveals a complex regulatory landscape in which jurisdictions are adopting a range of legal approaches, including criminal punishments, civil remedies, or a combination of methods. We also find that legislators are frequently turning to tort-law frameworks to address the harms of deepfakes. This article explores the current landscape of tort-based regulations of deepfakes. In addition to providing an overview of the most recent legislative developments, we unpack and compare the various tort-law methods arising at the state and federal level. We further consider how lawmakers are modifying existing tort laws to address the unique concerns raised by deepfakes. While individualistic tort remedies allow victims of deepfakes to seek direct recourse through familiar private rights of action, our analysis also identifies practical and conceptual limitations with this approach. Traditional tort frameworks struggle to address key challenges posed by deepfakes, including anonymous creation, viral distribution at technological scale, and harms affecting both individuals and society broadly. In light of these limitations, legislators are innovatively adapting traditional tort concepts — such as standing, mental states, causation, immunities, and remedies — to address deepfakes’ unique characteristics. Yet the very need for these adaptations reveals some of tort law’s shortcomings and suggests a space for complementary regulatory approaches. We consider some potential approaches that could provide this more complete framework, like tort liability for entities that enable deepfake creation and circulation, and civil enforcement mechanisms that empower state actors to vindicate both individual and societal interests. Ultimately, our findings suggest that while tort law provides a valuable foundation for addressing deepfake injuries, comprehensive protection will require additional regulatory frameworks beyond traditional tort remedies.

The combination of technological advancement of generative AI with the ever-increasing importance of participation and presence in the digital world for one’s overall quality of life makes the harm from deep fakes a pressing problem. There is a need to both regulate deep fakes and remedy the harm they cause. In the paper, I focus on tort law, privacy-based response to a subset of the deep fake problem: nonconsensual intimate fakes (‘intimate fakes’). I make the following claims in the paper: (1) Those who distribute intimate fakes of identifiable real persons where the image appears to be real should be liable to the plaintiff for breach of their privacy. Those who make these images (with a self-use exception) and those who host and view them should be likewise liable, but I will focus on distributors here. This follows from the position that false private information, even if defamatory, implicates the plaintiff’s privacy interest and should allow the plaintiff to sue for breach of privacy for reasons both pragmatic and conceptual (including an analogy I make between information and genome). This position is largely endorsed by existing English law. (2) There should (and probably could) be no tort liability for the distribution of intimate fakes of fictional characters not reasonably understood to refer to a real person. However, liability for breach of sexual privacy manifested in intimate fakes should extend also to cases of look-alikes (including of an invented character) and in this sense as well liability should be strict. (3) A strong case exists for imposing liability also in cases in which the intimate fake is a known fake: ostensibly fake, or accompanied by disclaimer that it is. I first explain why these known fakes merit liability, highlighting what makes their distribution both harmful and (civilly) wrongful. I then explain that liability could be conceptualised as (also) vindicating the claimant’s privacy interest. (4) There should be no tort liability for the creation of an intimate fake for self-use. However, the creator should be strictly liable for the distribution of the intimate fake; moreover, there is much to support liability of the creator towards the subject of the image should the latter learn that the image exists even if the intimate fake itself was not distributed.

This Article explores how U.S. tort law can respond more effectively to the distinct harms posed by deepfakes, including reputational injury, identity appropriation, and emotional distress. Traditional tort doctrines, such as defamation, the right of publicity, and intentional infliction of emotional distress (IIED), remain fragmented and ill-suited to the speed, scale, and anonymity of deepfake dissemination. Using a comparative functionalist approach, the Article analyzes how China and Singapore respond to deepfake harms through structurally divergent but functionally instructive frameworks. China’s model combines codified personality rights with intermediary obligations under a civil law regime, while Singapore adopts a hybrid approach that integrates common law torts with targeted statutory and administrative interventions. Although neither model is directly replicable in the United States, both offer valuable comparative insights to guide the reform of U.S. tort law. The article advances an integrated governance model for U.S. tort law: reconstructing personality-based torts, repositioning tort law through conditional intermediary liability, and clarifying constitutionally grounded limits for speech-based claims. Drawing on Chinese and Singaporean legal approaches, the Article sets out a comparative reform framework that enables U.S. tort law to better address deepfake harms while safeguarding autonomy and dignity in AI-driven digital environments.