Geneva convenes the first UN Global Dialogue on AI Governance this week. The draft road map for the next twenty years of digital governance contains the word gender zero times. Here is our week.
The treaties of the next decade will not be signed in Geneva this week. But their vocabulary is being chosen. For the first time, the United Nations convenes a Global Dialogue on AI Governance, sharing one week and one building at Palexpo with the WSIS+20 High-Level Event and AI for Good. Whatever language survives this week travels into the road maps, and road maps have a way of hardening into the world.
So here is a fact about the current draft of the Joint Implementation Road Map, the document meant to steer the information society for the next twenty years. The word gender appears in it zero times.
A missing word sounds like a grievance. But it is a data failure, and data failures are not abstract. Medicine spent a century treating the male body as the human default, and women still pay for it in misdiagnosed heart attacks and drugs dosed for bodies not theirs. Machine learning is a mirror with a long memory: train a system on that record and it learns the default, then applies it at scale, in triage software and risk scores and eligibility algorithms, with the serene confidence of arithmetic. Meanwhile the authorship gap compounds the data gap. On AI patents worldwide, women appear as inventors at a rate of sixteen percent, lower than their share across all fields of invention. Absent from the data, absent from the authorship, and now, if the draft stands, absent from the measurement framework of the next two decades. Nothing gets governed that has not first been counted.
Counting, as it happens, is my assignment at the inaugural Global Dialogue on AI itself.
On Monday afternoon I moderate the closing panel of the opening day’s session on AI opportunities and implications, with a lineup that spans the whole machine: Jian Wang, who founded Alibaba Cloud; Deemah AlYahya, Secretary-General of the Digital Cooperation Organization; Kitty van der Heijden, Deputy Executive Director of UNICEF; Ambassador Philip Thigo of Kenya, who negotiated AI into the Global Digital Compact and drove the first UN resolution on AI’s environmental footprint; Bilal Mateen of the UN’s new Scientific Panel on AI; and Ambassador Guilherme de Aguiar Patriota of Brazil, chairing the working group now drafting the UN’s report on global data governance. After a day of principles, we take up the machinery: what honest measurement of AI’s environmental cost would require, what human oversight looks like when it is real rather than recited, and whether any of it can interoperate across borders and unequal capacities.
The rest of the week runs the same argument upstream. Monday at nine we start where governance either happens or quietly doesn’t: procurement. Every safeguard not written into a contract is a safeguard that does not exist, and governments are signing AI contracts at scale without the tools to write them well. With Cambridge and the Council of Europe we put an operational framework on the table.
Monday afternoon, with the Gender in Digital Coalition, we bring the indicators the road map is missing, while it can still be amended.
Tuesday the argument reaches its sharpest point, in a session called Fit for Whom: AI systems are being deployed on women without being validated on women, and we have the receipts, three research papers tracing the cascade from male-default data to skewed outcomes in health, justice and social care.
By Tuesday evening it becomes a question of binding law, with the CEDAW machinery and the Chair of the UN Working Group on Discrimination Against Women and Girls.
Wednesday the argument goes inside the companies. At the Palais des Nations, OHCHR’s B-Tech Project and BSR launch a new workstream on human rights in AI product design, with Google, AWS, Microsoft, Proton, Infosys and NASSCOM at the table. My brief there is the question this whole week keeps circling: what would due diligence guidance have to contain to catch gender and other differential impacts at the design stage, rather than arriving as a generic checklist after the harm has shipped. Principles most often fail in the product spec.
Thursday it goes behind closed doors. I facilitate the Leadership Circle Lab of the Network of Women Ministers and Leaders in ICT, built deliberately as the opposite of a panel: small groups, no audience, no communiqué theater. The question on the table is the one polite sessions never quite reach: where exactly does a commitment to gender-responsive AI lose traction in your system, at the budget line, the procurement rule, the standards table, the appointment, and what single lever would change it. Ministers are rarely asked to be that concrete in public. That is why the door is closed.
Friday morning the week splits across two continents. In Geneva, I serve as High-Level Track Facilitator of the WSIS Forum’s LeadersTalkX on digital and AI in partnership, half an hour in Room C with leaders spanning the African Union, national governments, industry and academia, on what multistakeholder cooperation has to become if any of this is to hold.
And in Seoul, at the same hour, HumRights-Bench will be presented at the International Conference on Machine Learning (ICML) AI4Law Workshop: the first expert-validated benchmark testing whether AI models can reason about human rights law. The models perform near chance. Their sharpest failure is the first step, recognizing whether a human rights question is present at all.
So the zero in the road map is not something we will be protesting from the corridor. Between the facilitator’s chair, the moderator’s chair, and the closed room of ministers, the pen is in the room, and so are we.
Women at the Table does all of this with a tiny team and no funding currently attached to any of it, which is either absurd or clarifying, depending on the day. If the argument matters to you, come to a session, send this to someone who holds a pen on a road map, or write to me. The rooms are below.
Where to find us:
Sunday 5 July, 14:00–17:00 · Feminist AI Innovation from the Global South: From Community to Global Governance Commitments · Club de la Presse, Geneva · with the Inclusive AI Research Network and the Gender in Digital Coalition · in person only
Monday 6 July, 09:00–09:45 · Public Trust and AI Procurement: From Principles to Practice · WSIS+20, ITU Room L1 · with the University of Cambridge and the Council of Europe ·
Monday 6 July, 14:00–14:45 · Measuring What Matters: Embedding Gender Equality in AI Governance Through Gender-Specific Indicators · WSIS+20, ITU · with Derechos Digitales, APC, Equality Now and IT for Change ·
Monday 6 July, 16:36 · Global Dialogue on AI Governance, Thematic Cluster 1 · Palexpo, Breakout Room A · the panel described above, moderated by Caitlin Kraft-Buchman · livestreamed on UN Web TV
Tuesday 7 July, 10:00–10:45 · Fit for Whom? Sex-Stratified Data and the Integrity of High-Risk AI · WSIS+20, ITU Room H2 · with FemTechnology, UNDP and Ambassador Muhammadou Kah of The Gambia ·
Tuesday 7 July, 16:00–16:45 · Enhancing CEDAW Accountability through AI Governance and Tech Value Chains · ITU Room E · organized by CEDAW, with OHCHR and UN Women
Wednesday 8 July, 09:00–10:30 · B-Tech 2.0: AI Product Design Workstream Launch · OHCHR B-Tech Project and BSR · Palais des Nations, Room XI · with Google, AWS, Microsoft, Proton, Infosys, NASSCOM, OECD, Research ICT Africa and InternetLab · by invitation
Thursday 9 July, 11:00–11:45 · Gender-Responsive AI Standards: From Rhetoric to Reality · Leadership Circle Lab, Network of Women Ministers and Leaders in ICT · Palexpo Room K · by invitation
Friday 10 July, 10:00–10:30 · LeadersTalkX: Co-Creating Tomorrow, Digital and AI in Partnership · WSIS Forum High-Level Track · Palexpo Room C · facilitated by Caitlin Kraft-Buchman as High-Level Track Facilitator · on-site and virtual ·
Friday 10 July · HumRights-Bench at the AI4Law Workshop, ICML 2026, Seoul ·
https://betterimagesofai.org / https://creativecommons.org/licenses/by/4.0/