Wikimedia Foundation Artificial Intelligence and Machine Learning Human Rights Impact Assessment

Introduction

At the Wikimedia Foundation, we believe that access to knowledge is a human right. Our mission is to ensure everyone, everywhere can access and share reliable information freely and openly on Wikipedia and other Wikimedia projects. Access to free and open knowledge, supported by the fundamental right to freedom of expression, empowers people to exercise many other rights enshrined in the Universal Declaration of Human Rights, including the rights to education, artistic expression, economic advancement, and political participation.

Wikimedia projects and Wikimedia volunteers everywhere occupy a unique space in today’s online information ecosystem. This ecosystem is, however, rapidly evolving. The introduction and rapid advancement of emerging technologies such as large language models (LLMs) and other kinds of generative artificial intelligence (AI) introduce opportunities as well as challenges related to the creation, access, and distribution of information. Generative AI is fundamentally changing how the public seeks, receives, and imparts information and ideas online, raising novel questions about the role and responsibility of the Foundation and Wikimedia volunteer communities in this ecosystem.

AI and machine learning (ML), however, are neither new to Wikimedia projects nor to the Wikimedians who make them possible. Both the Foundation and volunteer communities have developed numerous ML tools to support Wikimedians in contributing, editing, and curating the ever-growing volume of knowledge across the projects as far back as 2010. Several of these tools have harnessed ML and AI to assist volunteers with frequently recurring tasks such as identifying vandalism or flagging when citations are needed. Most tools currently used were developed before the introduction of generative AI.

About this AI/ML HRIA report

This human rights impact assessment (HRIA) forms an important part of the Wikimedia Foundation’s long-term efforts to meet the commitments articulated in our Human Rights Policy. It follows and builds upon our 2020 organizational Human Rights Impact Assessment (which identified five categories of significant human rights risks facing the Foundation and Wikimedia volunteer communities) and our 2023 Child Rights Impact Assessment (which studied the unique risks facing children on Wikimedia projects). Given that generative AI was not widely available or used when these HRIAs were completed, this report considers how emerging technologies may affect the human rights risks that had been identified previously and also what new risks have emerged.

The Foundation commissioned Taraaz Research, a specialized research and advocacy organization working at the intersection of technology and human rights, to independently conduct the assessment and write this report. The assessment was conducted between October 2023 and August 2024 and offers suggested areas for research, investigation, policy, and evolution. The purpose of the impact assessment was to identify and analyze potential impacts, opportunities, and risks emanating from the use of AI and ML technologies in the Wikimedia ecosystem. The report does not necessarily represent the views of the Foundation or volunteer communities.

In researching this report, Taraaz examined specific AI tools developed by the Foundation to assist volunteers as well as known social and human rights issues that could be exacerbated by AI, such as bias, discrimination, and knowledge equity. The report identified their potential associated risks and proposed concrete recommendations that the Foundation and volunteer communities could implement to mitigate the risks in question so as to maximize the positive impact that the tools can have on our free and open knowledge mission. It does not reflect actual observed risks, harms, opportunities, or benefits that have resulted from the use of ML or AI technologies on Wikimedia projects. As a part of this impact assessment, Taraaz consulted Foundation staff, individual volunteers, volunteer affiliates, civil society organizations, and external subject matter experts.

After reviewing the initial report, the Foundation and Taraaz partnered to carry out a comprehensive review in order to prepare a version for publication. Content that could potentially expose volunteers, either individuals or communities, to risk, or that could possibly empower malicious actors to exploit identified vulnerabilities, has been redacted in order to safeguard them. This report is, therefore, a redacted version of the original.

What does the report include?

This impact assessment report includes:

A foreword, providing additional context for the HRIA, why it was commissioned, and how the Foundation has responded since receiving the report;
An executive summary, which summarizes the scope and methodology, salient risks of in-house ML-enabled products and recommendations for the products in question, and salient marginal rights risks of generative AI for Wikimedia projects and volunteer communities as well as recommendations regarding its use in the WIkimedia context;
A detailed six-part analysis framework, which includes threat identification, existing risks and defenses, evidence of marginal risk, defense strategies, and underlying assumptions, and that focuses on three categories of machine learning-enabled products: in-house tools developed by Foundation staff for Wikimedians’ content editing and moderation support; generative AI and its potential marginal human rights risks in the Wikimedia context; and Wikimedia's core projects’ data used for external ML development.

In order to improve the impact and accessibility of this report, the foreword and the executive summary have been translated into Arabic, Chinese, French, and Spanish. These texts are available as PDFs linked on this page.

What work has the Foundation done to respond to the report’s findings and recommendations since it was submitted?

It will take time to socialize the findings and recommendations of this report, discuss them with volunteers, and decide how best to work together to implement them, but we believe that is the best way forward that is consistent with our mission and values.

Some recommendations, however, can be implemented by the Foundation directly. We have already begun to lay the groundwork for developing and using AI technologies in ways that protect and uphold human rights in service of our free knowledge mission. The Foundation's teams working on this HRIA and these AI technologies collaborated to identify which recommendations would be the most impactful and could feasibly be implemented in the short-term.

For example, based on recommendations provided by this HRIA, we have decided to prioritize improving our existing methodology for evaluating the potential human rights impact of AI/ML models and products. We will begin including the findings of these evaluations in the model cards that the Foundation publishes for the AI models and products it creates. In the future, AI/ML model cards and project pages will describe the potential human rights impacts of a given model and AI/ML-powered feature, as well as steps that the Foundation will take to reduce the potential negative impacts. This will further improve transparency around our AI models by informing Wikimedia volunteers and the general public about the potential human rights impacts we identify when developing these models and how we plan to mitigate those risks.

Importantly, the Wikimedia Foundation released a three-year strategy about the use of AI to empower volunteer editors in April 2025. The strategy focuses on the importance of using AI in ways that enhance the work of Wikimedians, while offering volunteers the ability to choose between AI-assisted or traditional workflows. The strategy recommends doing so by:

Supporting Wikipedia’s moderators and patrollers with optional AI-assisted workflows that automate tedious tasks in support of knowledge integrity;
Giving Wikipedia’s volunteer editors time back by improving the discoverability of information on Wikipedia to leave more time for human deliberation, judgment, and consensus building;
Helping volunteer editors share local perspectives or context by offering translation automation and adaptation of common topics;
Scaling the onboarding of new Wikipedia volunteers with guided mentorship.

Looking forward (September 2025)

As we explained in the blog post announcing the Foundation’s AI strategy:

Our efforts will use our long-held values, principles, and policies (like privacy and human rights) as a compass: we will take a human-centered approach and will prioritize human agency; we will prioritize using open-source or open-weight AI; we will prioritize transparency; and we will take a nuanced approach to multilinguality, a fundamental part of Wikipedia.

The publication of this HRIA report represents a starting point for reflection about the potential societal impacts that will emerge from the interplay between the Wikimedia projects, Wikimedia volunteers, the Wikimedia Foundation, and AI in the coming years. Such emerging technologies are, relatively speaking, in their early stages, and their full potential and risks are only beginning to be discussed and understood. New applications, opportunities, and risks will inevitably emerge. This HRIA report establishes a baseline understanding of these opportunities and risks so that we can chart a timely path forward together.

How can I provide feedback on this report, or tell the Foundation about how my community is working with AI?

Over the coming months, we will create opportunities to hear directly from your communities and you about the findings and recommendations of this report, as well as your perspectives on the opportunities and risks associated with AI and ML in the Wikimedia ecosystem. You can already leave your thoughts and comments on this Talk page or register to join us at one of the following conversations on this topic:

Come join us so that we can make meaningful decisions together about how these tools can be used for the common good as well as continue to make sure that knowledge is human!