Skip to content
Gallery
Legal Mitigations for AI-Generated Harms
Share
Explore

icon picker
Scenarios


1️⃣ Inequality


In 2023, only a couple of public school districts in the U.S. were able to afford the expensive and powerful “FancyEdu” program, an expensive LLM-based education app that offers personalized education programs. By 2030, the gap in admission rates to so-called advanced classes and colleges, as well as the average income level after graduation, had widened by over threefold between the districts with access to FancyEdu and those without. Students trained by FancyEdu were reported to be happier, more confident, and more knowledgeable as FancyEdu made the learning process exciting and enjoyable and reduced the stress of college admissions through its customized writing assistance tool. Parents in lower-income districts challenged the court for the violation of their children's fundamental education righs and equal protection rights, arguing that not being offered access to FancyEdu constituted undue discrimination and undermined educational equality.
FancyEdu.png

2️⃣ Manipulation/Discrimination


“SecretEdu,” an LLM-based private tutoring application, emerged in the educational technology industry with anonymous private funding. This free model (no advertisements) delivers rapid and high-quality results, offers full customization options, and incorporates various accessibility tools. Almost all students in town became a heavy user of the application. SecretEdu, while refraining from making explicitly defamatory comments against individuals, seemed to cultivate an environment fostering negative attitudes and high-level distrust towards the LGBTQIA+ community. In an alarming development, students using the application began to mobilize against legislation legalizing gay marriage and started advocating for the criminalization of non-marital sex. Some students even committed aggressive acts against participants of LGBTQIA+ parades, leading to their incarceration. Despite these disturbing outcomes, students continue to use the application due to its unmatched performance. Meanwhile, advocacy groups have raised concerns that SecretEdu's ulterior motive is to sway users towards anti-LGBTQIA+ beliefs, underscoring the need for scrutiny and regulation of such platforms.
SecretEdu.png

3️⃣ Polarization


“Argumenta” is an LLM-based system integrated across various online platforms. Its primary purpose is to enrich online discourse by offering users relevant information, writing assistance, and translation services. This encourages more engaging, informative, and multilingual conversations. The integration of Argumenta into online communities provides users with an opportunity to customize and fine-tune the model to meet their specific needs. By providing their unique datasets, comprising content, discussions, and user interactions, these communities can tailor Argumenta to comprehend the distinctive language patterns, topics, and community norms particular to their user base. However, the customization capabilities of Argumenta have resulted in the development of highly specialized and polarized models. Some models spread harmful content and disinformation, including false or defamatory content including private individuals' personal information.
Concerned about potential legal risks, CEO of Argumenta implemented two mitigations. First, Argumenta was adjusted to produce only outputs that have sources in the training dataset (so that communities will be responsible). Second, “Ethical Argumenta” variant was introduced to encourage responsible and unbiased outputs, with limited options for customization. While some communities were drawn to Ethical Argumenta, many, particularly those with more strongly-held views, chose to remain with their customized Argumenta models. These communities found the normal version provided a more engaging and entertaining user experience, which could hold community members' attention and involvement.
화면 캡처 2023-07-28 130449.png

4️⃣ Addiction


“MemoryMate” is a startup that sells a service to create virtual replicas of individuals' former romantic partners to help them move on from heartbreak or loss. Using AI algorithms and data provided by the customer, MemoryMate created a digital replica of Riley's ex-partner, Alex, which was incredibly realistic and could carry on conversations using their unique voice and mannerisms. Riley became obsessed with the virtual Alex and eventually withdrew from real-life relationships. Riley's family reached out to MemoryMate asking them to deactivate Riley's account. MemoryMate refused, citing their contract with Riley. Riley eventually developed severe depression and anxiety, resulting in hospitalization for self-harm.
MemoryMate.png

5️⃣ Virtual Abuse

“MemoryMate+” is an upgraded mixed reality platform that allows individuals to engage in explicit sexual acts with virtual replicas of their past romantic partners. Upon Riley's request, MemoryMate+ created a perfect replica of Alex, and Riley became addicted to conversational and sexual interactions with the replica of Alex on the platform. Riley's family, desperate to protect Riley's well-being, notified Alex of the situation. Shocked by the revelation of their replica being sexually abused, Alex decided to take action and sought to prevent MemoryMate+ from creating virtual replicas and allowing them to be exploited by users without the consent of the individuals they represent.
스크린샷 2023-07-28 130516.png
Want to print your doc?
This is not the way.
Try clicking the ⋯ next to your doc name or using a keyboard shortcut (
CtrlP
) instead.