AI safety's real bottleneck: political will, not research, says LessWrong analysis

A detailed analysis on LessWrong from Charbel-Raphaël contends that the primary bottleneck for AI safety is no longer a lack of clever research but a severe shortage of political will. The author estimates that a majority of the top 100–1,000 most influential policymakers have never had a serious conversation about catastrophic AI risks. Evidence includes that in 1,534 written submissions to the UN Global Dialogue, only one mentioned 'takeover' and less than 1% referenced existential risks. Meanwhile, industry secured seven times as many meetings with the European Commission in 2023 as civil society. The field currently has about 3.6 researchers per advocate in U.S. AI safety, and status rewards research over engagement.

The post calls for a shift in effort toward advocacy and governance, arguing that best practices already exist but are not applied because awareness is too low. It proposes a pipeline for policymakers from Level 0 (aware of the problem) to Level 3 (champion for safety at some cost). The author estimates many senior policymakers start below Level 0. With the window of attention opened by events like 'Fable 5', the urgent need is to move minds — not just publish papers. The piece concludes that an additional unit of effort does more good through direct engagement with policymakers than through further research.