The main building of Humboldt University Berlin, where the seminar took place, 15-16 June 2026
On 15 and 16 June 2026, I took part in “Standard-Takers and Standard-Makers: Global Practices and Politics in the Age of AI,” a workshop hosted by Humboldt University Berlin bringing together researchers, standards bodies, and policy practitioners to examine technical standardization as a strategic arena, not a purely technical one. The workshop materials are available here.
The first day set the frame: standards are never neutral. They shape interoperability, market access, security, competitiveness, and technological sovereignty, and in fast moving fields like AI, data governance, and cybersecurity, early participation in standard setting can quietly determine future regulatory pathways. A recurring theme was the divide between “standard-makers” and “standard-takers.” Speakers from African and Global South perspectives pointed to the barriers developing countries face, limited resources, weak institutional capacity, thin technical expertise, and little real influence in global forums, and argued for stronger regional coordination and coalition building. Discussions also touched on the fragmentation caused by the growing number of formal bodies, industry consortia, proprietary platforms, and powerful technology companies competing to set the rules.
The second day turned to how countries, firms, and institutions actually seek influence in global standardization. National strategies were compared, including how developing economies are building quality infrastructure, stakeholder coordination, and harmonization with ISO and IEC work to move beyond simply taking standards set elsewhere. Other sessions looked at how domestic AI capability, patents, publications, and institutional leadership translate into authority inside international standards bodies, and at the growing role of Chinese institutions in ISO related knowledge production as part of a broader shift in standardization power. An afternoon panel turned to AI in practice, covering data management certification as a foundation for AI innovation and private sector cooperation on frontier AI safety.
I spoke on the closing panel, which asked whether global standards can actually reflect local values, alongside discussions of digital public infrastructure as a possible “third way” and digital product passports as a case study in keeping systems interoperable across divergent regulatory traditions. My own talk, “Human Rights as Guardrails in AI Standardization,” presented ITU as a consensus based UN platform where governments, industry, academia, and civil society negotiate technical recommendations together.
Technical standards are not neutral. They shape how technologies impact people’s rights, from privacy and dignity to equality and non-discrimination.
I walked through where this stands in practice at ITU. WTSA-24 adopted Resolution 101 on trustworthy AI standardization and Resolution 105, which referenced human rights in metaverse standards for the first time, while the Freedom Online Coalition, representing 42 states, called for inclusive, rights based standards development, a direction echoed in the ISO/IEC/ITU Seoul Joint Statement. I also shared results from an internal ITU-T survey of 55 responses across all study groups, TSAG, and all six ITU regions: 84 percent consider human rights relevant to their work, but only 56 percent say their study group actually considers them, only 33 percent point to any existing process or guidance, and 40 percent report low familiarity with the concepts, alongside strong demand for good practices from other standards bodies, targeted training, and AI powered review tools.
A comparative look across other standards organizations showed both promising practices and shared struggles. IEEE’s Ethically Aligned Design and UNECE’s Gender-Responsive Standards show how ethical principles can be embedded directly into technical work. ETSI is introducing human rights checks into its process, and at IETF security considerations are mandatory for every standard while human rights considerations remain voluntary and handled case by case. CEN/CENELEC gives consumer, labor, and environmental groups a formal voice in its processes. Across the board, though, the same obstacles keep showing up: fragmented approaches, the difficulty of translating rights language into technical language, low awareness among technical communities, and no shared tools for assessing human rights impact.
On the current work I am leading at ITU-T, I presented a draft lightweight guideline for Study Groups built around fifteen screening questions designed to catch risks early, plus a standard human rights consideration paragraph technical committees can adapt to their own recommendations, alongside a training track: a 90 minute masterclass, a self paced ITU Academy course with case studies and a certificate, and an instructor led course applying the tools directly to ITU-T use cases. To make the idea of “translation” concrete, I walked through a few examples: facial recognition and remote biometric identification, where the relevant safeguards include accuracy testing across demographic groups, strict access control, and human oversight, smart cities and IoT, where local language access and data minimization matter most, and digital humans, avatars, and synthetic media, where consent mechanisms and age appropriate design are the key levers.
The screening questions I find most useful in practice are simple ones: who could be affected by a draft Recommendation, including children, older persons, people with disabilities, rural communities, and users with limited connectivity? Is data collection actually limited to what is necessary? Would affected users understand when and how a technology affects them? And are there real mechanisms to detect, report, and correct errors, misuse, or harm? None of these questions require a law degree to answer, which is rather the point: the goal is to give engineers a practical entry point into human rights, not to turn every study group meeting into a legal seminar.