The Human Rights Center at UC Berkeley School of Law has released a novel human rights assessment of the use of large language models (LLMs) in the fields of law, journalism, and education.
“An International Analysis of the Human Rights Impacts of Large Language Models” was conceived to answer how the use of LLMs may create human rights risks and opportunities when used within the fields of law, journalism, and education; how human rights risks and opportunities associated with the use of LLMs vary according to geographic and socioeconomic contexts; and what measures can be implemented by which entities to effectively mitigate human rights risks and leverage the opportunities presented by LLMs.
“What surprised us the most was how people in different parts of the world use LLMs so differently in their work, with practitioners in Global North states predominantly much more risk adverse. While this obviously led to varying human rights risks, we learned about new opportunities different uses posed, particularly in certain Global South states,” said Betsy Popken, executive director of the Human Rights Center and author of the report alongside researchers Vyoma Raman and Camille Chabot.
Drawing on the human rights impact assessment framework used as a due diligence tool under the UN Guiding Principles for Business and Human Rights, researchers blended stakeholder interviews with a literature review to identify rights-impacting trends. They interviewed 56 people from 24 countries—including at least one expert and one practitioner in each sector on each continent—and representatives from corporations developing LLMs, between 2023 and 2025. They also analyzed existing sector-specific model evaluations to connect performance metrics to human rights impacts in the fields of law, journalism, and education.
Researchers found cross-cutting risks to freedom of thought where overreliance on LLMs can stifle higher-level critical thinking. They also found that cultural and ideological homogenization challenges global adoption to using LLMs, creating risks to the right to equality and nondiscrimination. These impacts vary widely based on usage, resulting from disparate resources, access, and training.
LLMs were found to improve potential access to justice by assisting litigants who do not have counsel, such as in small claims settings. However, overreliance in high-risk settings were found to risk due process and fair trial rights, and the right to remedy, by providing inaccurate or unnuanced advice.
In journalism, LLMs were found to provide assistance with writing and research that can improve the productivity and labor conditions of journalists operating in resource-constrained settings. However, hallucinations were noted as contributing to the spread of misinformation.
LLM usage was found to positively and negatively impact the right to education. Personalized learning systems made it easier for educators to adequately assist students with their individual challenges, but the alignment of LLMs to educational standards and the ability of LLMs to help with student recall remains a challenge.
Researchers made a suite of recommendations to LLM developers in their report, including diversifying datasets to better represent global communities; designing refusal mechanisms and response tones based on the severity of human rights risks; and developing profession-specific LLMs to address field-specific challenges. They also recommended establishing clear remediation processes to identify, assign responsibility for, and remedy harms resulting from LLMs; requiring comprehensive human rights risk assessments across the AI lifecycle; and creating regulations that target specific high-risk use cases, such as judges using LLMs to make their rulings.
This report comes on the heels of Assessing Human Rights Risks in AI: A Framework for Model Evaluation, published in arXiv and featured in the AAAI/ACM Conference on AI, Ethics, and Society (AIES) 2025, where co-author Rahman, in collaboration with co-authors Popken and Chabot, developed an approach to evaluating a model to make grounded claims about the level of risk a model poses to particular human rights. Co-authors Betsy Popken and Vyoma Raman also published the piece “Building LLMs? Think Beyond Borders” with Tech Policy Press, to explore why this work is so important.
…
Press Contact: Maggie Andresen [mandresen@berkeley.edu]