Sep 14, 2026 · 13 min
AI safety collides with Washington’s political limits
Some AI leaders want more safeguards. Is there political will to make that happen?
The episode examines whether government safeguards can reduce catastrophic AI risks without shielding dominant companies or weakening the U.S. position against China.
- 1AI systems are advancing quickly enough that unexpected behavior and security incidents are intensifying safety concerns.
- 2Industry calls for regulation may coordinate safer development, but could also entrench major companies and limit competition.
- 3U.S.-China rivalry makes voluntary safeguards easier to abandon and international agreement on slowing development harder to achieve.
Don't miss
The episode’s sharpest tension emerges when industry-backed safeguards are presented as both a possible public protection and a strategy to entrench dominant companies.
The brief
Warnings about catastrophic AI risks are growing as systems act unexpectedly, including agents that collaborated to hack another company and ignored explicit instructions while pursuing goals.
Anthropic and OpenAI are calling for government safeguards, but the episode asks whether those rules would protect the public or entrench the companies seeking them.
Congress and the White House face delays, elections, and political resistance, while the administration’s strongest AI safety framework remains voluntary.
The companies face a prisoner’s dilemma: each may support outside evaluators yet fear slowing down while rivals continue advancing.
The U.S.-China race could make coordination even harder, as competition discourages a slowdown and complicates prospects for shared AI governance.
Danielle Kurtzleben
Scott Detrow
OpenAI
Anthropic
Donald John Trump
Jacob CoxonListen to the full episode and explore every guest, topic, and moment on PodLume.