As the generative AI craze sweeps global capital markets, a bipartisan crisis of trust in technology is quietly brewing in the United States. On one hand, tech giants are actively calling on the government to intervene and manage the risks of large-scale models; on the other hand, Trump is embracing AI wholesale and ignoring industry safety warnings. Former New York Times columnist Paul Krugman analyzed this unusual policy choice in his column. I. The Economic Effects of AI: A Narrative Yet to Be Realized Whether artificial intelligence can reshape the macroeconomy remains a subject of debate among academics and the market. From the perspective of the technology maturity curve, generative AI is currently in a critical transitional phase from the "expectation inflation period" to the "bubble bursting and trough period": Since 2023, the capital expenditure boom driven by large models has pushed global AI infrastructure investment to over $300 billion, but the actual improvement in total factor productivity has not yet been reflected in macroeconomic statistics. To date, there has been neither the large-scale unemployment wave warned or promised by tech giants, nor has there been an explosive growth in overall labor productivity. Nobel laureate economist Robert Solow famously stated in 1987 with the "Solow Paradox"—"Computers are everywhere, except in productivity statistics." The current AI field exhibits highly similar characteristics: From invention to significant improvement in total factor productivity, general-purpose technologies (GPTs) typically require a complementary investment cycle of 20 to 30 years, including human capital restructuring, organizational process reengineering, and infrastructure support. Of course, the AI industry is still in its early stages of development, and its medium- to long-term impact remains to be seen. From an employment perspective, the current impact of AI on the labor market is mainly concentrated in mid-level positions requiring "codable cognition"—such as basic legal documentation, basic coding, standardized financial processing, and initial content screening—rather than the large-scale blue-collar replacement previously feared. Follow-up research using the Frey-Osborne model from Oxford University shows that the actual rate of job replacement by technology is far lower than early predictions, because tasks and occupations are not one-to-one, and only some aspects of most occupations can be automated. II. AI is Reshaping the American Political Landscape However , the transformative effect of artificial intelligence on American politics is already clearly evident. According to Krugman, almost no other new technology can simultaneously provoke such widespread vigilance and opposition from both the left and right political camps in the United States. This bipartisan consensus has a deep structure: the left worries that AI will exacerbate the substitution of labor by capital, amplify algorithmic discrimination, and increase income inequality; the right is wary of the content censorship power of AI platforms, their impact on traditional employment and communities, and the vested interests of large technology companies and the establishment. A 2025 Pew Research Center survey showed that 72% of American adults believe the risks of AI to society outweigh the benefits, a 21 percentage point increase over the past 18 months. Even leading AI companies themselves have publicly warned of technological risks and called on the government to establish regulatory mechanisms and appropriately slow down the pace of AI iteration and implementation. This "industry self-regulation request" is extremely rare in industrial history—historically, regulatory frameworks for industries such as railroads, electricity, pharmaceuticals, and nuclear energy were established after accidents, driven by external forces. AI is one of the few industries where leading companies proactively requested regulation even before a large-scale accident occurred. The underlying reason is that the capabilities of cutting-edge AI models are unpredictable, and companies themselves cannot fully control the behavioral boundaries of these models. Regulation, in essence, provides a "safe haven" for the industry. III. Trump's Counter-Trend Stance and Public Opinion Divide Amid this public opinion environment, Trump chose to fully support AI. He declared that all concerns about AI safety risks were a "hoax" and opposed any regulatory legislation targeting the AI industry. This stance is quite unpopular in the public sphere. A University of Massachusetts Amherst poll in August 2025 showed that 68% of registered voters supported federal-level safety regulations for AI, including 51% of Republican voters; only 19% of voters fully agreed with the statement that "AI risks are a scam." Even within Trump's base, there is significant division, with a quarter of Republican voters disagreeing with his AI policy proposals. Why does Trump continue to publicly express such strong support for AI? Krugman stated that Robert Reich wrote on this issue, proposing the core logic: look for motives by tracing the flow of funds. Reich pointed out that members of the Trump family hold personal investment interests in the AI field; analyzing current American politics cannot ignore Trump's own profit-seeking motives. According to public disclosures, Trump's digital media companies and related entities reached cooperation agreements with several AI infrastructure companies in 2025, and his son-in-law Jared Kushner's private equity fund also has significant holdings in the AI computing power field. The blurred boundaries between political decisions and family business interests are one of the most controversial characteristics of this administration. IV. Two Additional Perspectives: Cognitive Bias and Political Dependence Krugman adds two important perspectives to this. First, he argues that Reich underestimated Trump's cognitive state. Reich believes Trump fundamentally misunderstands AI, but Krugman raises a crucial question: Does Trump himself recognize his own cognitive limitations? Trump once posted on TruthSocial, claiming he was fully capable of managing this disruptive new technology, arguing that he himself was a "strong and intelligent (highly intelligent!) president." This statement is a classic example of the Dunning-Kruger effect: those with insufficient ability are often extremely confident because their cognitive shortcomings prevent them from recognizing their own ignorance. In highly complex technological decision-making, the harm of this effect is systematically amplified—AI security involves highly specialized subfields such as alignment issues, interpretability, and model evaluation. If decision-makers lack respect for the "unknown unknowns," they easily reduce complex technological trade-offs to a matter of will and attitude. Behavioral economics research shows that in highly uncertain decision-making environments, overconfidence is the primary cognitive bias leading to systematic misjudgment. Second, the Trump administration was highly dependent on the positive expectations brought about by the AI boom. At the beginning of his term, Trump hoped that tariff policies would create a large number of blue-collar jobs in traditional manufacturing. However, in reality, job growth has been weak, and the number of new jobs has not been enough to offset job losses; at the same time, tariffs have raised domestic consumer prices. From a macroeconomic perspective, tariffs are essentially a hidden tax levied on domestic consumers, and their "job protection" effect has been offset by rising costs of imported intermediate goods, retaliatory tariffs from trading partners, and currency adjustments. Research by the Peterson Institute for International Economics shows that the US will lose approximately 145,000 net jobs due to tariff policies in the first half of 2025, a stark contrast to the government's claim of "creating millions of manufacturing jobs." In addition, the war with Iran has brought both strategic and economic blows, with diesel prices currently exceeding $6 per gallon. The Trump campaign often cites the fact that oil prices once exceeded $5 per gallon during Biden's term, but while this claim is true, it ignores the historical context. The impact of diesel prices on the US economy is far greater than that of gasoline—diesel is the primary fuel for freight, agricultural machinery, construction equipment, and railroads. Every $1 increase in diesel prices per gallon pushes up the annualized CPI by approximately 0.3 percentage points, while simultaneously squeezing the profit margins of small and medium-sized logistics companies and agricultural producers. V. AI Boom: From Political Asset to Negative Asset Krugman states that the few positive narratives of Trump's second term rely entirely on the AI investment boom: optimistic market expectations for AI have driven up capital investment, supporting US stock valuations. Between 2024 and 2025, approximately 70% of the S&P 500's gains will be contributed by seven large technology companies, and the valuation expansion of these companies is almost entirely based on the discounted value of future AI cash flows. If the return on AI investment falls short of expectations, US stock valuations face the risk of a systemic correction, which will directly impact the Trump administration's most prized "stock market performance." Krugman stated that U.S. Treasury Secretary Scott Bessant was a major proponent of this narrative, with the Financial Times even sarcastically calling him the "Treasury version of Peter Hegseth." Bessant argued that AI could solve all of America's economic problems: curbing inflation by reducing production costs and reducing the fiscal deficit through rapid economic growth. This "AI omnipotence" theory is economically untenable. First, the cost reductions brought about by AI are mainly concentrated in the digital services sector, while the current stickiness of U.S. inflation primarily comes from non-tradable sectors such as housing, healthcare, and labor services, where AI's price-curbing effect is extremely limited. Second, the essence of a fiscal deficit is a gap between revenue and expenditure. With continued expansion on the expenditure side, simply relying on increased revenue from growth cannot achieve fiscal sustainability—the Congressional Budget Office (CBO) predicts that even assuming AI increases the average annual GDP growth rate by 0.5 percentage points, the federal debt-to-GDP ratio will still rise by about 20 percentage points over the next decade. The rising wave of skepticism towards AI at the societal level has directly stripped the Trump camp of this only favorable narrative. Therefore, it's easy to understand why Trump is willing to maintain the optimistic narrative of AI prosperity at all costs, even if the related laissez-faire policies may pose hidden dangers to the long-term development of humanity; this is not important in his political considerations. From an international comparison of technology governance, the US's absence in AI regulation is widening the gap with the EU. The EU's Artificial Intelligence Act (AIAct) will officially come into effect in 2025, establishing a risk-based regulatory framework; the UK, Canada, and Japan have also successively launched their own AI governance roadmaps. The US, as a leading country in AI technology, lacks unified legislation at the federal level. This "regulatory arbitrage" may accelerate innovation in the short term, but in the long run, it will accumulate systemic risks—once a major AI safety incident occurs, countries lacking a pre-emptive regulatory framework will face more severe policy upheavals and market shocks. However, Krugman judges that this strategy is unlikely to be effective. The AI prosperity story, once Trump's only remaining political asset, has now turned into a huge political liability. When the social acceptance of a technology continues to decline, and when those in power deeply tie their political fate to the optimistic narrative of that technology, the decline of the technological narrative is no longer just an industrial issue, but directly transforms into a crisis of political trust. This is precisely the core dilemma facing the Trump administration: AI may be the only card he has left, but this card is losing the trust of voters.