UN AI for Good 2026

Greetings from Geneva!

Adrianna and Niall are both in Geneva this week for UN AI for Good as well as for World Summit on the Information Society.

Lots to report, including:

• Maria Ressa, Yoshua Bengio and the rest of the UN Independent Scientific Panel on Artificial Intelligence launched their preliminary report
• UN Secretary-General António Guterres appealed for far-reaching global controls on AI, and reiterated a continued support for 'no killer robots'
• Luisa Machado of Equilab raised the important point: why are global lawmakers more interested in regulating children and young people's relationship to technology, than in regulating technology companies' relationship to power?

One key finding of the UN Independent Scientific Panel preliminary report:

• That AI today doesn't comprehensively perform as well, or as safely, in many lower resource languages
• The report gave the example of AI translating "smallpox as syphilis, gonorrhea as diabetes", and mixing up "antibiotics" with "insecticide" in Tigrinya, a language spoken by 7 - 9 million people in Eritrea and northern Ethiopia

Multilingual and multicultural performance, safety and security are core components of what we do here at Future Ethics.

Representatives of many governments prepared statements, stating their concerns and positions on AI.

• Many African and Latin American nations, including Botswana, Rwanda and Guatemala were concerned about AI widening the digital divide
• Spain, France, Switzerland and others expressed concern about regulation not catching up with the pace of AI development, as well as concern for the environment, misuse of AI for misinformation and surveillance

Meanwhile, Adrianna participated in two sessions:

• On 6 July, 5pm, she was part of a WomenInGenAI panel at WSIS on "Responsible AI in Public Services: From Policy Principles to Deployable Workflows"

• Adrianna shared her experience building software in the public sector, then developing TEVV (testing, evaluation, validation and verification) expertise on AI in the public sector
• Along with other panelists, Adrianna reiterated that pilots that fail in production potentially fail when use cases and fitness for use are not properly sized
  • "AI is just one of many tools in our digital toolbox in the public sector. Build services that work for the people, don't build services just to use AI."
• Adrianna discussed what it's like running AI red teams and evaluations for state, federal and international governments as well as healthcare organizations
  • Why people don't run as many red teams as they should (cost, time, resources required)
  • But she expressed a belief that testing costs will come down, and talent / resources to perform these tests will increase
  • What's most important is to develop an internal sense of 'we should test AI before we deploy it to the public'

Tomorrow, on 8 July, Adrianna will speak at AI for Good.

Details of event

Date: Wednesday, 8 July 2026, 14:30 – 17:40 (local time)
Venue: ITU AI for Good Global Summit 2026 venue Palexpo Room V

The workshop will focus on the following key topics:

• Technological frontiers and future trends of foundation models
• Evaluation methodologies, indicator frameworks, and global benchmarking for foundation models
• Balancing foundation models evaluation capability and responsible AI development
• Standardization and international collaboration pathways for foundation model evaluation
• Co-developing foundation models evaluation systems and international initiatives under the AI for Good vision

This workshop is an open, exploratory discussion aimed at facilitating exchange and cooperation among all parties on the standardization of foundation model evaluation.

Come say hi if you are also there!