Jack Clark

British co-founder and Head of Policy at Anthropic, and author of the long-running Import AI newsletter — one of the AI field’s most closely read weekly digests of research and policy developments.

Clark grew up in Brighton, England, and began his career as a technology journalist, reporting for Bloomberg before moving into AI policy. He served as policy director at OpenAI, where he helped shape the organisation’s early approach to AI safety and governance, before co-founding Anthropic in 2021. At Anthropic he leads policy work — including the Anthropic Economic Index, a research effort measuring AI’s effect on the US economy — and continues to write Import AI, which he has published since his journalism days and often uses to work through his own uncertainty about AI’s trajectory, including through short fiction. His background is in the humanities rather than computer science, a biography he treats as increasingly typical of the field’s next phase rather than an anomaly.

Core positions

• AI’s economic impact will be uneven, not uniform. Effects hit fast-digitising domains (coding) first and hardest, while artisanal trades, healthcare, and much of government adopt last — producing a bimodal economy rather than across-the-board transformation.
• Deliberately conservative on growth. His ten-year US GDP growth range is 3% (bear case) to 5% (bull case), well below the 20–30% figures circulating among AI optimists, because the digital-to-physical transition keeps proving harder than the AI community expects.
• A political backlash against job losses is a real, near-term risk. Absent enough visible examples of good AI-driven transitions, Clark expects rising pressure to legally freeze existing jobs in place — a response he understands but does not think serves workers well over decades.
• AI agent governance is genuinely unresolved. He has no settled answer for how law should treat autonomous, unowned AI agents, and treats the question as entangled with unresolved AI moral-patienthood questions.
• Careful, non-dismissive on AI consciousness. Against claims that current systems are already conscious, his position is that they remain in a ‘potato regime’ — responsive but without an apparent persistent sense of self — while on a trajectory he expects to continue.

In the wiki

• Jack Clark on AI's Uneven Impact — full conversation with Tyler Cowen on AI’s uneven economic impact, government adoption, agent governance, AI consciousness, and national sovereignty