Session
Description:
Generative AI systems are becoming embedded in critical information and communications infrastructure. As they increasingly intermediate speech and the flow of information, human rights governance frameworks must be built into their development and deployment — protecting users' rights to free expression, privacy, and freedom from discrimination. Companies developing and deploying generative AI remain behind in operationalizing human rights due diligence for their outputs. Few have published clear policies articulating how human rights frameworks are considered or implemented within their models and operations. Developers don't need to start from scratch, however. Existing guidance is adaptable to this context, particularly through a multistakeholder process. Stakeholders can draw on past collaborations that strengthened human rights protections in content moderation: in October 2021, a coalition of civil society organizations released the second iteration of the Santa Clara Principles on Transparency and Accountability in Content Moderation, giving social media companies concrete guidance for respecting human rights in moderation. The UN Guiding Principles on Business and Human Rights offer foundational guidance for assessing business impacts on human rights and building processes to address risks. UNESCO's Guidelines for the Governance of Digital Platforms outline how to address digital technology's risks in line with international human rights law. Translating these frameworks for generative AI is now needed. This session will convene a multistakeholder group to workshop how transparency, accountability, and human rights due diligence can be adapted from content moderation and incorporated into generative AI's development and deployment. Participants will work toward concrete recommendations — using the Santa Clara Principles as a starting template — for what a "Santa Clara Principles for AI" might include, focused on incorporating human rights principles into model outputs. Other urgent considerations — AI's impact on labor markets, the environment, and affected communities — are important but beyond this workshop's scope.
Generative AI systems are becoming embedded in critical information and communications infrastructure. As they increasingly intermediate speech and the flow of information, human rights governance frameworks must be built into their development and deployment — protecting users' rights to free expression, privacy, and freedom from discrimination. Companies developing and deploying generative AI remain behind in operationalizing human rights due diligence for their outputs. Few have published clear policies articulating how human rights frameworks are considered or implemented within their models and operations. Developers don't need to start from scratch, however. Existing guidance is adaptable to this context, particularly through a multistakeholder process. Stakeholders can draw on past collaborations that strengthened human rights protections in content moderation: in October 2021, a coalition of civil society organizations released the second iteration of the Santa Clara Principles on Transparency and Accountability in Content Moderation, giving social media companies concrete guidance for respecting human rights in moderation. The UN Guiding Principles on Business and Human Rights offer foundational guidance for assessing business impacts on human rights and building processes to address risks. UNESCO's Guidelines for the Governance of Digital Platforms outline how to address digital technology's risks in line with international human rights law. Translating these frameworks for generative AI is now needed. This session will convene a multistakeholder group to workshop how transparency, accountability, and human rights due diligence can be adapted from content moderation and incorporated into generative AI's development and deployment. Participants will work toward concrete recommendations — using the Santa Clara Principles as a starting template — for what a "Santa Clara Principles for AI" might include, focused on incorporating human rights principles into model outputs. Other urgent considerations — AI's impact on labor markets, the environment, and affected communities — are important but beyond this workshop's scope.
Policy Question(s)
A. What provides meaningful transparency across the AI value chain?
B. Where are human rights protections weakest? What would close gaps across jurisdictions?
C. How can frameworks integrate economic, social, and cultural rights?
D. What would encourage human rights due diligence, and what barriers exist?
E. How do AI models differ from social media? Have dual-service companies bridged that gap? Are Santa Clara Principles lessons transferable?
F. What accountability and appeal mechanisms fit AI systems?
G. What role should local/regional partners play in developing recommendations for protecting rights in AI model development and deployment?
Expected Outcomes
Building on the policy questions, the session will produce: (1) a comparative mapping of transparency and accountability gaps between content moderation and generative AI systems; (2) an outline of what a human rights due diligence framework for generative AI companies should include, grounded in the UN Guiding Principles on Business and Human Rights and (3) concrete next steps for the development of a civil-society informed human rights, transparency, and accountability framework for generative AI. These outputs will inform ongoing civil society advocacy toward generative AI developers, feed into broader multistakeholder AI governance discussions (including relevant UN processes), and lay groundwork for a possible follow-on drafting initiative modeled on the original Santa Clara Principles process.
Format
Roundtable
Duration (minutes): 90
Format description: We want to encourage sharing ideas and initiatives from online and onsite attendees. A roundtable format enables equal visibility and speaking opportunity for all participants, rather than a hierarchy of speakers and listeners. The session will open with brief remarks from civil society representatives and human rights experts familiar with implementing human rights due diligence in the social media and AI contexts, followed by a facilitated breakout discussion in which participants workshop draft recommendations in small, moderated groups. Groups will reconvene to report back and refine a shared set of takeaways. This format was chosen because the goal of the session is not simply to present a problem, but to collectively draft the building blocks of a rights-based due diligence framework — mirroring the multistakeholder drafting process that originally produced the Santa Clara Principles. A 90-minute session allows sufficient time for framing, small-group work, and report-back without losing participant engagement.
Duration (minutes): 90
Format description: We want to encourage sharing ideas and initiatives from online and onsite attendees. A roundtable format enables equal visibility and speaking opportunity for all participants, rather than a hierarchy of speakers and listeners. The session will open with brief remarks from civil society representatives and human rights experts familiar with implementing human rights due diligence in the social media and AI contexts, followed by a facilitated breakout discussion in which participants workshop draft recommendations in small, moderated groups. Groups will reconvene to report back and refine a shared set of takeaways. This format was chosen because the goal of the session is not simply to present a problem, but to collectively draft the building blocks of a rights-based due diligence framework — mirroring the multistakeholder drafting process that originally produced the Santa Clara Principles. A 90-minute session allows sufficient time for framing, small-group work, and report-back without losing participant engagement.
Hybrid Format: We will ensure one rapporteur is designated to monitor the online chat so we do not miss any contributions from online participants, alongside a separate online facilitator who can unmute or call on remote attendees at natural pauses. One breakout group will be online only and will report out directly to onsite attendees to ensure equal weight of contribution. One presenter will join online so their research is included regardless of location. All slides and pre-read materials will be shared with online attendees in advance, so they aren't at a disadvantage from being unable to glance at onsite materials in real time. We will use shared note-taking infrastructure so everyone can access notes during and after the event.
Organizer 1: Kate Ruane, Center for Democracy and Technology
Organizer 2: Shumaila Hussain Shahani, Tech Global Institute
Organizer 2: Shumaila Hussain Shahani, Tech Global Institute
Speaker 1: Komaitis Konstantinos , Civil Society, Western European and Others Group (WEOG)
Speaker 2: Alexandria Walden, Private Sector, Western European and Others Group (WEOG)
Speaker 3: Thobekile Adebayo , Civil Society, African Group
Speaker 4: David Greene, Civil Society, Western European and Others Group (WEOG)
Speaker 5: Elonnai Hickok, Civil Society, Western European and Others Group (WEOG)
Speaker 2: Alexandria Walden, Private Sector, Western European and Others Group (WEOG)
Speaker 3: Thobekile Adebayo , Civil Society, African Group
Speaker 4: David Greene, Civil Society, Western European and Others Group (WEOG)
Speaker 5: Elonnai Hickok, Civil Society, Western European and Others Group (WEOG)
Moderator
Shumaila Hussain Shahani, Civil Society, Asia-Pacific Group
Online Moderator
Kate Ruane, Civil Society, Western European and Others Group (WEOG)
Rapporteur
Kate Ruane, Civil Society, Western European and Others Group (WEOG)
What will participants gain from attending this session?
