Anthropic’s Jack Clark takes the AI ‘sudden change’ warning to the IMF
The Anthropic co-founder will deliver the Per Jacobsson Lecture in Bangkok, arguing that AI progress is driving massive and abrupt shifts
Jack Clark, co-founder and head of policy at Anthropic, is scheduled to deliver the Per Jacobsson Lecture on October 17, 2026, during the IMF/World Bank Annual Meetings in Bangkok, Thailand.
What Clark plans to say
The Per Jacobsson Foundation hosts the event. This year’s lecture will focus on how fast-moving AI could reshape three areas: scientific discovery, the economy, and everyday life.
Clark is expected to argue that the evolution of AI capabilities is producing abrupt shifts rather than gradual ones.
Clark will likely point to Anthropic’s own experience as evidence. As of August 2026, the company’s Claude model accounted for 26% of its R&D activities. In February of the same year, that figure was less than 1%.
A month of warnings before the main event
Throughout September 2026, Clark spoke with various media outlets about the implications of AI technologies, raising concerns about the real-world risks that certain AI systems pose and calling for stronger regulatory measures.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
Two proposals stood out. The first is verifiable oversight, meaning regulators and the public should be able to confirm that AI developers are actually following the rules. The second is mandatory safety protocols such as kill switches, a reliable way to shut down an AI system if it starts behaving in ways nobody intended.
Why the IMF stage matters
The IMF/World Bank Annual Meetings will include sessions dedicated to AI’s impact on jobs and productivity.
What this means for policymakers and the AI industry
The most important audience for this lecture is not AI researchers. It is the officials who set fiscal and monetary policy for entire countries.
Those officials care about two questions above all: what happens to productivity, and what happens to jobs. A sharp rise in AI’s share of R&D work — from less than 1% to 26% at Anthropic in roughly six months — hints at large productivity gains, but also raises obvious questions about how human roles change when software takes on more of the load.
For the AI industry, Clark’s regulatory proposals of verifiable oversight and mandatory kill switches would impose real obligations on developers, including Anthropic itself.