Safe & Ethical AI Vienna: Slow Down AI? Or "WHOEVER WINS AI, WINS"?
This is the fifth meetup of Safe & Ethical AI Vienna -- a cross-disciplinary community of professionals from academia, industry and policy who work on, think about and care deeply about the risks posed by AI, be they ethical, societal, psychological or existential.
The topic of this meetup is:
Slow Down AI? Or "WHOEVER WINS AI, WINS"?
For years, calls to slow down advanced AI development have faced a familiar objection: the benefits are too great, competition is too intense, and if one company or country stops, another will simply take its place. But what happens when calls for slowing down come from the people leading the race?
In September, Anthropic CEO Dario Amodei published “We Must Pace the Frontier”, arguing that safety work needs more time to catch up with advances in AI capabilities. His proposal includes independent evaluators working inside AI labs (apparently they are arranging that Accenture will be auditing them), shared safety standards, and international coordination. Sam Altman and Elon Musk endorsed the call for pacing within hours. Earlier in the summer, OpenAI and Anthropic had also temporarily paused particular training and testing activities following safety incidents.
However, a temporary pause to fix a security problem is one thing; a coordinated effort to slow the growth of AI capabilities is another. Amodei explicitly says that "pacing" does not mean halting technical progress. The idea is to make further advances depend on our ability to understand, test and safeguard the systems we are building.
Not everyone welcomed the call: the day after Amodei's essay, President Trump rejected it, saying that the US is leading China in AI and that he wants to keep it that way, because "whoever wins AI, wins". The White House AI adviser David Sacks responded that the labs are free to slow down on their own, but should not use safety concerns to push for regulation that suits them. At the same time, Amodei's own proposal relies on the race with China, since he argues that democracies must keep as large a lead as possible in order to have room to slow down at all.
At our last meetup, we discussed the OpenAI–Hugging Face incident and what it might tell us about losing control of AI. This time, we turn to the response. Further disclosures, growing public attention and proposals from within the industry have made the question more urgent: how fast should we be developing increasingly powerful AI, and who should get to decide?
Would slowing down give us time to make AI safer? And what, specifically, should we do with that time? Could independent oversight make companies' commitments credible? Does the competition with China make slowing down impossible, or is the lead exactly what makes it affordable? How could cooperation work between competing companies and countries, and can it work without the support of their governments? And could rules intended to protect the public end up concentrating power in the hands of a few large labs?
We'll begin with a short overview of recent developments, followed by small-group discussions and a shared conversation. We'll consider the arguments for slowing down, the objections, and what a workable proposal might look like.
You don't need to support a slowdown—or have a technical background—to join. Come and think it through with us.
Some background reading:
“We Must Pace the Frontier” — Dario Amodei’s case for slowing capability advances and strengthening oversight
“Pacing model development in an era of cyber-critical capabilities” — OpenAI’s account of its temporary slowdown and changes to safeguards
“A Framework for Frontier AI and the Dawning of a New Age” — Demis Hassabis’s proposal for a frontier AI standards body
The vision for Safe & Ethical AI Vienna is to build:
A cross-disciplinary community spanning research, industry and policy
A space to exchange ideas and develop collaborations
A connection to the local and international AI safety landscape, including events, opportunities and new research
A stronger presence for Vienna in the global AI safety conversation
If you think AI poses some of the most important and underexplored risks of our time, and you’d like to be around others who feel the same — come along.