ABBA-AI
| Reference number | |
| Coordinator | RISE Research Institutes of Sweden AB - Samverkande Autonoma System |
| Funding from Vinnova | SEK 149 479 |
| Project duration | January 2026 - April 2026 |
| Status | Ongoing |
| Venture | Advanced digitalization - Industrial needs-driven innovation |
| Call | Collaborations with the US in AI, digital infrastructure and cyber security |
Purpose and goal
ABBA-AI (Autonomy Bound By Assurance, a transatlantic framework for risk-reducing applied AI) builds a transatlantic bridge for safe autonomy in cyber-physical systems—from drones to humanoid robots. It targets the “Assurance Gap” and “autonomy creep,” where AI capabilities expand without matched accountability, traceability and authority. Via a strategic Silicon Valley presence (incl. NVIDIA GTC), the project forms partnerships that connect US AI performance with Swedish safety architectures.
Expected effects and result
Deliverables: gap analysis of standards, policy and regulation in autonomy governance, and a Step-2 consortium for minimum framework of safe autonomy.
Planned approach and implementation
The approach follows three phases: (1) Preparations in Sweden: mapping of humanoid and dual-use actors, definition of “assurance controls” against autonomy creep (identity, shutdown, insurance, risk recursion) and pitch material. (2) Implementation in the US: NVIDIA GTC, targeted meetings with AI/humanoid and defense autonomy actors and workshops. (3) Follow-up: synthesis, gap analysis report, LoI/Joint Concept Note and consortium for stage 2.