AI Companies Can't Regulate Themselves. They Should Regulate Each Other
AI companies face a collective action problem where competition undermines safety investments, leading to a race to the bottom. A proposed solution is to adapt the financial industry's model of federally supervised self-regulatory organizations (SROs) to AI, allowing industry-written rules under government oversight. This approach could address challenges like information asymmetry, rapid technological change, and the need for preemptive risk mitigation.
- ▪Competition among AI labs is driving a race to the bottom on safety, with companies like Anthropic and OpenAI scaling back safety measures to keep pace.
- ▪The Frontier Model Forum already coordinates risk management among major AI labs but lacks statutory authority, mandatory membership, and government oversight.
- ▪Financial regulators have long used federally supervised self-regulatory organizations (SROs), such as FINRA, to enable industry self-governance under government supervision.
- ▪AI regulation must overcome four challenges: competitive pressures, information asymmetry, the rapid pace of technological change, and the need for ex ante intervention to prevent irreversible harms.
- ▪An effective AI regulatory body would need the ability to update rules quickly, access proprietary technical information, and enforce binding safety standards across all frontier labs.
Opening excerpt (first ~120 words) tap to expand
Congress Cybersecurity & Tech Democracy & Elections AI Companies Can’t Regulate Themselves. They Should Regulate Each Other. Mark Thomas Wednesday, April 29, 2026, 10:10 AM Share On: Share on Facebook Share on X Share on LinkedIn Share on BlueSky Share on Threads Print this article Adapting a long-standing institutional model from financial regulation would let the industry write binding safety rules under government oversight. (https://closebot.com/blog/different-ai-providers-and-how-their-responses-differ-with-same-prompting/; CC BY-NC 4.0, https://creativecommons.org/licenses/by-nc/4.0/) Mark Thomas Meet The Authors Subscribe to Lawfare Competition is preventing artificial intelligence (AI) safety.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at Default.
Discussion
0 commentsMore from Default
