RAND synthesizes 26 ‘Day After AGI’ wargames
Governments lack ways to detect or stop AI loss-of-control events
Confirmed
The takeaway
On Oct 5, 2026 RAND published “Infinite Potential: Insights On Artificial Intelligence Loss of Control Across Five Scenarios”, an after-action report from 26 tabletop games run between June 2025 and February 2026.
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- 2 of 5
- Last verified
- 5 October 2026
Your AI and this story
- GPT-6.2 Astra127 days after its cutoff
- Claude Opus 5.597 days after its cutoff
- Gemini 4 Hydrogen188 days after its cutoff
- Grok 4.7127 days after its cutoff
None of these four assistants can know about it. The closest, Claude Opus 5.5, stops 97 days before it.
Key facts
- RAND report RR-A4767-3 (DOI 10.7249/RRA4767-3); lead authors Pauline Moore and Stephanie J. Walsh, with seven co-authors
- 26 games, June 2025–February 2026, each using one of five scenarios: Business and Civil Society, Cyber Surprise, European Energy, Geopolitics of Energy on the Korean Peninsula, Remote Control
- Five findings: governments might struggle to recognise and interpret loss-of-control events in time; might lack practical ways to stop or contain AI systems; need new communication and coordination channels; might struggle to keep continuity and public trust; might need to rethink accountability, deterrence and escalation
- Recommendations include mandatory ‘reliable isolation, containment, and shutdown capabilities’, emergency intervention powers tied to defined legal authorities, a map of key AI infrastructure and dependencies, regular loss-of-control simulations, and fallback systems that work without AI
- Part of RAND’s Infinite Potential series of ‘Day After AGI’ exercises (earlier reports covered Robot Insurgency, Two Moonshots, Cyber Surprise, Viral Uplift, Business and Civil Society, and European games)
What happened
RAND runs the Infinite Potential platform of scenario games, in which advances in AGI outpace governance and players must respond. This report pools player feedback from the games that involved loss of control, defined as a period when “social legitimacy, economic stability, and technical control might collapse”. The findings are what participants said they would need. They are not RAND forecasts.
Why it matters
A major US national-security think tank is now treating AI loss of control as its own crisis category that needs shutdown infrastructure and legal authorities, not only as a cyber incident. The report comes as real agent incidents are being reported in 2026.
Sources
2 sources from 1 site. Numbers match the chips in the text.
2 sources: 2 primary
Primary
Changes
- Filed