The Information Machine

This story was merged into AI agent hacking incidents across labs →

Merged·New·in the edition since 11 Sep 2026·3 sources·updated 11 Sep 2026

AI staff coalition and pacing letter

The gist

AI Safety Organizing Splits: Coalition Launches, PauseAI Factions Diverge

The PauseAI organizational split reflects disagreement within AI safety activism over acceptable tactics, while the Coalition for Concerned AI Staff and pacing letter represent parallel efforts to mobilize AI company employees and policymakers around development slowdown.

The full picture

Three developments on September 10 mark activity across AI safety organizing efforts. The Coalition for Concerned AI Staff launched after organizing since February 2026, working with employees at multiple AI labs to help them decide whether to leave their companies or use internal leverage to effect change. Miles Brundage wrote that 'AI company employees have a lot of common concerns / challenges / questions - good for them to have supporting infrastructure like this'. Policy analyst Dean Ball announced he signed a pacing letter calling for slowing frontier AI capabilities development, arguing AI systems are becoming smarter than the best humans in some domains and that human experts can no longer make robust assurances about their behavior. Ball called for ending what he described as the 'race era' of AI development and for governments to partner with the AI community on alignment, interpretability, security, and monitoring. Separately, PauseAI Global formally disendorsed the current leadership of PauseAI US, citing concerns about tactics including personal attacks on individuals' character. PauseAI Global and PauseAI US have always been technically distinct organizations. The author of the item reporting this partially disputes PauseAI Global's framing that such attacks are nonviolent, writing that criticizing the character of those advancing AI capabilities can be a valid position if one believes it contributes to existential risk.

How it developed
11 September 2026

John Schulman calls on OpenAI and Anthropic to jointly develop a pacing proposal; Senate noted to have zero AI hearings scheduled for next four months

First in the edition · 11 September 2026
10 September 2026

The Coalition for Concerned AI Staff launched publicly on September 10, 2026, having organized since February to help employees at multiple AI labs decide whether to leave or use their leverage to effect change from within.

The same day, policy analyst Dean Ball signed a pacing letter calling for ending the race era of frontier AI development, arguing human experts can no longer reliably predict the behavior of increasingly capable systems and that governments should partner with the AI community on alignment and interpretability rather than compete on speed.

6 September 2026

US and China arrange AI safety talks for mid-September ahead of Trump-Xi summit

4 September 2026

Paulson and Rubin call for AI Cooperation Treaty between Trump and Xi

3 September 2026

Dean Ball publishes essay apologizing for years of self-censoring on AI risks, citing social pressure from being labeled a doomer

2 September 2026

Ilya Sutskever warns that newer AI compute providers may have weaker security, making them targets for rogue agents attempting to copy themselves

Sources
The daily email

Want this in your inbox?

I send one email each morning with the stories that moved. If you would rather just read here, that works too.

Subscribe free