AI Slowdown Calls Test Whether Society Has a Seat at the Table

Situation

The scenario describes growing calls from researchers associated with OpenAI and Anthropic to slow AI development amid warnings of potentially catastrophic harm. These warnings put the companies’ development pace, release decisions, and safety governance under scrutiny because the consequences could extend far beyond customers, employees, and investors. The warnings warrant serious examination, but they do not establish misconduct, demonstrate that extinction is likely, or by themselves determine which restrictions would be effective.

4th Stakeholder assessment

The 4th Stakeholder question is whether society’s interests can meaningfully constrain frontier AI development—not merely be considered alongside commercial priorities. The scenario does not establish existing company practices that merit credit under the four pillars. A credible response would connect public benefit to the business engine, assign responsibility for downstream risks, prioritize consequential interventions, and give affected communities visibility and influence over decisions.

The framework, part by part

1. 1

   Doing More Good

   Building positive societal impact into the core business engine.

   Already aligned

   Nothing notable here yet.

   What to improve

   Make a documented public-benefit and safety case a required gate for each major model release, specifying intended societal benefits, evidence supporting them, safeguards, and post-release outcome measures.

   CEO and head of product

   This embeds societal value into the core development and deployment process rather than treating capability growth, customer uptake, or separate philanthropic work as sufficient evidence of public benefit.

   Related framework articles:

   • 3. Tuning your core business engine to do more ‘good’
   • 4. Making your business and social impact one and the same
2. 2

   First & Second-Order Impacts

   Owning downstream harm, not just direct effects.

   Already aligned

   Nothing notable here yet.

   What to improve

   Establish a jointly developed, independently tested frontier-risk protocol with peer labs and relevant public authorities, defining evaluation requirements and explicit thresholds for restricting, delaying, or pausing deployment.

   Chief safety officer, with board approval

   Misuse, cascading failures, and competitive pressure can create harms that no single company controls. Shared thresholds help address distributed responsibility without allowing collective action to replace each company’s own accountability.

   Related framework articles:

   • 9. Second-order business impacts defined
   • 11. Collective Action Step 1 - Mobilizing the collective
   • 12. Collective Action Step 2 - Mitigating the societal harm
3. 3

   Taking Bigger Swings

   Concentrating energy on the highest-impact priorities.

   Already aligned

   Nothing notable here yet.

   What to improve

   Adopt a board-reviewed Impact x Responsibility matrix that ranks safety interventions by potential societal harm reduction and company responsibility, makes uncertainty explicit, and assigns resources and deadlines to the highest-priority actions.

   Board risk committee

   Catastrophic warnings deserve rigorous evaluation without crowding out well-supported present harms. Explicit weighting helps distinguish high-leverage interventions from symbolic commitments and prevents commercial ROI from becoming the default safety prioritization rule.

   Related framework articles:

   • 13. The critical few vs. the trivial many
   • 15. Why we can’t use financial ROI as the prioritization metric for social impact
   • 16. A 4th Stakeholder prioritization metric matrix
4. 4

   Voice of Society

   Two-way dialogue, a published plan, and transparent progress.

   Already aligned

   Nothing notable here yet.

   What to improve

   Create a standing Voice of Society process that combines protected researcher reporting with input from independent experts and affected communities, and publishes a quarterly response plan showing concerns raised, decisions taken, accountable owners, and unresolved disagreements while protecting sensitive information.

   Board-appointed independent oversight lead

   Researchers’ public warnings are not evidence of a functioning company listening system. A traceable input-to-action process gives people outside the commercial stakeholder groups meaningful influence and makes company responses assessable.

   Related framework articles:

   • 21. A Voice of Society Engine Part 1: Listening Posts
   • 22. A Voice of Society Engine Part 2: A 4th Stakeholder Prioritization Group
   • 23. A Voice of Society Engine Part 3: Transparent Action