Anthropic’s Jack Clark tells the IMF and World Bank AI could push US GDP to $44T by 2030 — and wants ‘unusual and serious rules’

• Who: Jack Clark, Anthropic co-founder and head of public benefit, addressing the 2026 IMF–World Bank Annual Meetings in Bangkok on October 17, 2026.
• What: an internal Anthropic economic model forecasting potential US GDP of about $44 trillion by 2030, against a current level of just over $30 trillion.
• Three scenarios: Modest $34.1 trillion (1.6% growth), Substantial $36.3 trillion (8.3% growth), Extreme $44.4 trillion (32.4% boost) — the extreme case requires recursively self-improving AI.
• Timeline shift: Clark now expects full recursive self-improvement capabilities by 2027, an earlier estimate than before.
• The ask: international cooperation on ‘unusual and serious rules’ for AI; policymakers have ‘a small number of years’ to impose safeguards.
• Context: Reuters reports Anthropic is preparing a possible November IPO that could value the firm at up to $2 trillion.

Anthropic co-founder Jack Clark told hundreds of finance ministers and central bankers in Bangkok on Saturday that one of his company’s internal economic models forecasts AI could raise US output by almost 50% by 2030 — to about $44 trillion, from just over $30 trillion today — under conditions where the technology keeps improving. In the same lecture, he called for international cooperation to erect ‘unusual and serious rules’ for the technology, warning there is a shrinking window for policymakers to act, Reuters reported.

The $44 trillion figure, which Clark himself called ‘sounds very outlandish, I know,’ is the eye-catching ceiling of the model rather than its central estimate. His pitch to the room was simpler: if the technology keeps getting better — and he argued that AI is already vastly increasing scientific discovery — then ‘economic doctrine suggests that new industries and inventions would need to have been created to power such a major economic expansion.’

Three scenarios, one eye-catching ceiling

According to Crypto Briefing’s account of the model, which Anthropic released as an economic scenarios explorer in September, it lays out three paths to 2030: Modest, at $34.1 trillion (1.6% growth); Substantial, at $36.3 trillion (8.3% growth); and Extreme, at $44.4 trillion (a 32.4% boost). The extreme case assumes AI reaches high levels of autonomy — and, crucially, that recursively self-improving systems deliver the full effect.

That is where Clark moved the goalposts. He now expects full recursive self-improvement capabilities by 2027, an earlier estimate than he previously held — and exactly the capability the extreme scenario depends on. Public opinion, for its part, is not betting on the moonshot: surveys cited in the coverage line up most closely with the substantial scenario, with respondents estimating a 10% GDP boost alongside a 5% unemployment rate.

Wealth and peril on the same timeline

The forecast was the hook; the warning was the point. Clark told the room there is growing evidence of a ‘singularity’ where AI development outpaces human control, and that policymakers had ‘a small number of years’ to impose safeguards, according to AI Affairs’ reporting.

‘We should expect that in the coming years, the whole pace of progress of not just AI but science as a whole will speed up, and this will feel confusing and disorienting, and will bring with it great wealth and potentially great peril,’ he said. On the darker side of that ledger, he added: ‘Being able to use AI for science opens us up to the misuses of science as well, biological and cyber weapons and things we cannot yet imagine,’ per the Reuters account.

Jobs on both sides of the line

Clark was specific about who gets hit first. He said AI was already causing job losses in fields with ‘high AI exposure,’ including call centres and customer service — while enhancing employment in scientific research. What those gains net out to is an open question: the number of jobs the technology might create in entirely new industries remains, in his framing, a major unknown.

The timing is not accidental

Reuters noted that Clark’s remarks come as Anthropic prepares for an initial public offering expected in November that could value the firm at up to $2 trillion. His comments also echo chief executive Dario Amodei, who has pushed for greater coordination among AI companies as model capabilities accelerate. In September, Amodei and tech leaders from Google, Meta, SpaceX, Nvidia and OpenAI signed a White House accord calling for responsible training and development of frontier AI models.

Whether the $44 trillion lands or not, the message to the people who manage the world’s money was unambiguous: the models are getting better on a schedule that is beating expectations, and the rules are not. That is a combination Clark clearly thinks will not stay peaceful.