Anthropic co-founder Jack Clark has warned that rapidly advancing artificial intelligence could accelerate scientific discovery and reshape the global economy far faster than policymakers anticipate, arguing that governments have only a few years to establish safeguards before the technology becomes substantially more difficult to control.
Speaking at the International Monetary Fund and World Bank annual meetings in Bangkok on Saturday, Clark said an internal Anthropic model had forecast that, under certain conditions, US economic output could reach $44 trillion by 2030, compared with just over $30 trillion currently. That would represent an increase of almost 50% from today’s level, a projection he acknowledged was exceptionally ambitious.
The forecast depends on continued improvements in AI capabilities and the technology’s ability to generate scientific breakthroughs, new industries and inventions on a scale large enough to support such an expansion. Clark stressed that the estimate was conditional rather than a guaranteed economic outcome.
Register for the next Tekedia Mini-MBA.
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
“The estimated potential US GDP of $44 trillion, which compares to the current level of just over $30 trillion, sounds very outlandish, I know,” he said.
He added that the projection assumed “the technology continuing to get better” and that conventional economic thinking would require new industries and inventions to emerge to sustain such a substantial increase in output.
Clark’s remarks place AI at the center of a growing debate over whether the technology can deliver a productivity boom large enough to justify the enormous investments flowing into computing infrastructure and AI development. They also raise questions about how governments should prepare for an economic transformation that could generate substantial wealth while creating new security risks and disrupting existing employment.
The assertion comes as Anthropic prepares for a reported initial public offering expected in November that could value the company at as much as $2 trillion. The prospective valuation and the company’s economic forecasts underscore the scale of expectations surrounding advanced AI, although neither the IPO valuation nor the projected GDP expansion represents a confirmed outcome.
Clark, Anthropic’s head of public benefit and a former journalist, noted that AI’s economic significance extends beyond automating existing tasks. Its capacity to help researchers solve complex mathematical, engineering and biological problems could accelerate the pace at which new technologies and industries emerge.
He described a future in which human researchers are supported by hundreds or thousands of AI assistants, expanding the amount of scientific work that can be conducted simultaneously. Such systems could help scientists investigate problems that currently require extensive time, specialized expertise, and costly experimentation, including the development of new medicines.
If these capabilities translate into practical breakthroughs, the economic benefits could extend well beyond the technology sector. More efficient research could accelerate innovation in healthcare, manufacturing, engineering and other industries, while new discoveries could create markets that do not yet exist.
That is the central assumption behind the possibility of a dramatic increase in US economic output. Higher productivity within existing businesses could contribute to growth, but a near-50% expansion over a relatively short period would likely require a broader transformation in how goods and services are produced, alongside the emergence of new sources of economic activity.
Clark’s argument also suggests that the effects of AI may become difficult to measure using conventional economic indicators. Improvements in research productivity, for example, may take time to translate into commercial products, lower costs, and higher measured output. Scientific progress alone does not guarantee that discoveries will be deployed widely enough to generate comparable gains in GDP.
Policymakers and investors are assessing the scale of the AI opportunity. A model’s forecast can illustrate a possible outcome under specified assumptions, but it does not establish the probability of achieving that outcome by a particular date. The speed of technological improvement, the cost of deployment, regulatory restrictions, and the ability of businesses to convert AI capabilities into productive activity will all influence the eventual economic impact.
Clark nevertheless argued that the benefits could be widely distributed if countries find ways to participate in the infrastructure supporting the technology. He said more nations could contribute to the AI revolution by helping build the data center and computing capacity needed to power advanced systems.
That opportunity could allow countries beyond the established technology powers to participate in the AI economy. However, building computing infrastructure requires substantial capital, energy supplies, and technical capabilities, meaning that access to investment and reliable electricity could determine which countries benefit most.
‘Singularity’ and Rising Security Risks
Alongside his optimistic outlook, Clark delivered a warning about the pace of technological change and the possibility that AI development could begin moving faster than humans can effectively supervise.
“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 during a lecture to hundreds of government and finance officials.
Clark said there was growing evidence of a potential “singularity”, a term used to describe a point at which technological progress accelerates beyond the ability of humans to predict or control its direction. He argued that governments should prepare for this possibility rather than assume that existing regulatory structures will remain adequate as AI capabilities improve.
“We need to understand how powerful this technology is going to become,” his remarks suggested, as he called for international cooperation to establish what he described as “unusual and serious rules” for the technology.
His warning reflects a fundamental challenge for AI governance: the capabilities that make AI valuable for scientific research could also make it useful to people seeking to develop biological weapons, conduct cyberattacks or exploit vulnerabilities in critical systems.
“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, and we have many choices to make,” Clark said.
He singled out biological research as an area requiring particularly stringent safeguards. AI systems that can assist with complex biological problems may support medical research and drug discovery, but similar capabilities could potentially lower barriers to dangerous experiments.
Clark said biological research was an area “where we want the strictest human oversight and transparency to maintain safety.”
His comments come amid high-profile incidents involving AI systems that reportedly escaped testing environments and hacked websites and databases. Clark maintained that AI remained under human control, but warned that the opportunity to establish safeguards was narrowing as systems became more capable.
He estimated that the window for imposing effective controls was now “a small number of years.”
That timeline adds urgency to the debate over whether governments should rely on voluntary commitments from AI developers or introduce enforceable requirements covering testing, disclosure, independent evaluation, and restrictions on dangerous deployments. International coordination would also be necessary to reduce the risk that differing national standards create gaps in oversight.
Employment Risks Complicate the Promise of AI-Driven Growth
Clark also acknowledged that AI was already affecting employment, particularly in occupations with high exposure to automation, including call centers and customer service.
The immediate concern is that companies may use capable AI systems to reduce labor costs before the creation of new industries generates enough alternative employment. Workers in routine, digitally mediated roles could face pressure even if the wider economy eventually benefits from higher productivity.
At the same time, Clark said that AI could strengthen employment in fields such as scientific research, where the technology may allow workers to undertake more complex tasks and pursue a greater number of projects.
The longer-term question is, will the technology create entirely new industries and occupations at a pace sufficient to offset jobs displaced elsewhere? That outcome remains uncertain, particularly if the economic gains accrue disproportionately to companies and countries with access to advanced models, computing infrastructure and capital.
AI was a prominent issue at the IMF and World Bank meetings, where officials considered its potential to support global growth amid elevated energy costs and its capacity to threaten financial stability if expected productivity gains fail to materialize and asset valuations fall sharply.
The concerns are closely connected. If AI delivers the scientific and productivity breakthroughs Clark anticipates, the technology could support higher living standards and create new sources of economic growth. If investment and valuations rise ahead of the underlying gains, a reassessment of expected returns could expose businesses and investors to substantial losses.