Geneva Summit Session Outlines Capability-Based AI Standards and Exit Rules

Event report AIFOD Geneva Summit, August 2026. Published 1 November 2026.

Governments adopting artificial intelligence must ground procurements in realistic domestic capabilities rather than standalone technology purchases, according to discussions at the AIFOD Geneva Summit 2026. The session on cross-cutting standards, organised by the AI for Developing Countries Forum (AIFOD), which publishes this site, evaluated a model intended to embed technological sovereignty into public procurement from the outset.

Layered capabilities and exit strategies

The proposed framework establishes a three-tiered sequence: foundational infrastructure, enabling institutional capacity, and sector-specific applications such as healthcare, water, and agriculture. Delegates noted that advanced services cannot function without prerequisites including energy, connectivity, and digital public infrastructure. By mapping these dependencies, authorities in developing nations can assess whether the baseline infrastructure actually exists before attempting implementation.

A central focus of the discussion was provider lock-in. Participants argued that public bodies must mandate interoperable systems, exportable records, and transferable institutional procedures before signing agreements. As the session highlighted, "Exit is designed in, not negotiated later." Designing termination mechanisms from the start ensures that incoming suppliers do not rely on incumbent vendors to comprehend administrative operations.

Lifecycle oversight and pilot testing

The model organises public capabilities into four stages: setup, run, change, and retire. This lifecycle approach turns compliance into an ongoing monitoring process capable of incorporating local community evidence and evaluating broader social effects, such as impacts on childcare or unpaid labour. The framework is intended to produce adaptable blueprints that allow procedural lessons to travel internationally without forcing uniform implementation. The session outlined a proposed next phase to test the capability assessment in three to five pilot nations while defining sector-specific structures.