Infinite potential: business and civil society AGI exercises
B · Useful, limited A peer-reviewed RAND after-action report, but its content is a qualitative synthesis of five two-hour tabletop exercises with about 70 participants, with no measured data.
Summary based on the full text·Original: about 35 minutes to read
In short
An after-action report by Richard S. Girven, Beba Cibralic and Gregory Smith of RAND's Center for the Geopolitics of AGI, published April 2026. It synthesises five runs of a tabletop exercise in which business and civil society leaders, playing themselves, faced a fictional AGI release that triggered white-collar unemployment, then a cyber and infrastructure collapse with unclear attribution. Participants treated AGI as a social crisis as much as a technical one and concluded that preparedness means pre-authorised actions that work when human decision speed falls behind events.
Key numbers
- 5Day After AGI exercises run June to December 2025 in Washington, New York, London and San Francisco
- 70+participants (CEOs, investors, philanthropists, academics, civic leaders) playing themselves
- 7 / 7 / 6key issues, desired capabilities and playbooks distilled from the runs
As reported by the source.
English·Web page·opens rand.org
What it reports
- Turn 1 scenario: a fictional US firm releases an AGI-level model; GDP rises while white-collar unemployment and protests grow. Turn 2: autonomous-system malfunctions, data-centre failures and deepfakes degrade communications.
- Participants described loss of the internet as comparable in impact to a nuclear attack and asked for analog backups: radio relays, verified broadcast schedules, physical message distribution.
- "Decision window failure": decision cycles of months versus technology that moves in weeks or hours. Business participants admitted multiday board consultations would be unrealistic in such a crisis.
- Fiscal proposals: robot or algorithmic profit taxes, windfall taxes for workforce transition, global tax accords, and negative tax incentives for retaining or retraining staff. Participants split on whether heavy taxation would slow innovation.
- Coordination proposals: US-China AGI crisis hotlines, a joint government-industry operations cell, contingency plans for data-centre shutdown and restart.
- Seven desired capabilities, from analog communications and a public-private "AGI Emergency Action Framework" to compute governance and a registry of pre-cleared civil society organizations.
- The most recurring theme was a sense of helplessness once systems moved faster than leaders could understand them.
How they got it
Two-hour tabletop exercises with 10-16 participants each, led by a RAND facilitator, with verbatim notes on a common template. Notes were reviewed by the notetaker, a RAND analyst and a project lead, then compared across runs to extract convergent issues, capabilities and playbooks. Scenarios were set 12-18 months ahead and assumed 2025 institutions stayed in place.
Limits
- Findings reflect what invited participants said under a designed scenario, not observed behaviour or probabilities.
- Five runs and a US/UK audience; no counts of how many participants held each view.
- Funded from RAND operations and philanthropic gifts, including donors active in AI-risk funding.
Why it matters for companies
The report is a usable template for a corporate crisis exercise around AI-driven disruption. Its practical point for executives is speed: board and approval cycles built for quarterly change may not work if a model release shifts labour markets or a cyber event takes systems down for days. Pre-agreed triggers, contact lists and offline communication channels are low-cost preparations. The labour discussion also signals that employers may face pressure to fund retraining or transition schemes if displacement accelerates.
Summary based on the full text. Read the original for the full argument and data.Read the original on rand.org