AI Fears Are a “Hoax,” Trump Tells Nvidia CEO

President Donald Trump turned an unexpected phone call with Nvidia CEO Jensen Huang into an unusually direct statement of his AI agenda. On September 14, 2026, Huang was speaking at the All-In Summit when Trump called. Huang put the president on speakerphone, giving the audience an unplanned window into their conversation.

Trump praised data centers, insisted that the United States must beat China in AI, and dismissed fears about the technology by declaring that “the whole thing is a hoax.” The president specifically rejected the idea that robots or AI systems will take control, while suggesting that safety critics could be helping political opponents or China, whether intentionally or not. NBC News documented the exchange, including Huang’s more measured response after the call.

The comment is easy to treat as another unscripted Trump sound bite. It is more useful as a policy signal. In a few minutes, Trump tied together three ideas shaping his administration’s approach to AI: rapid deployment, massive data-center construction, and strategic competition with China.

He also blurred several different debates. Doubting a science-fiction scenario in which robots conquer humanity is not the same as showing that AI systems pose no safety, security, economic, or infrastructure risks.

Trump Turns a Speakerphone Call Into AI Policy

Huang initially answered the call without immediately identifying Trump to the audience. Once the president learned that the room could hear him, the conversation became a mixture of banter, praise for Nvidia, and policy messaging.

Trump told Huang that the United States needed to win the AI competition with China. He also called data centers “great” and argued that they could create wealth for people and states. That message closely matches Nvidia’s commercial interests. The company supplies many of the accelerators, networking products, and computing systems used to train and operate advanced AI models.

The president then moved from infrastructure to safety. His comments focused on the most extreme version of the AI-risk debate: autonomous machines taking control of the world. He dismissed that possibility and portrayed broader criticism as an obstacle to American development.

Huang did not directly confront Trump, but he also did not repeat the president’s central claim. After the call, the Nvidia CEO said he welcomed Trump’s sense of urgency and agreed that the country needed to build its AI capacity. However, Huang added, “I wouldn’t use the word hoax,” and noted that every major technology carries some risk.

That distinction matters because Huang had every reason to support the pro-development portion of Trump’s message. Nvidia benefits when technology companies and governments spend more on data centers. Even so, he stopped short of endorsing the idea that AI safety itself is fraudulent.

The exchange therefore revealed two overlapping but different positions. Trump framed safety concerns as politically motivated resistance to progress. Huang argued for faster development while leaving room for legitimate risk management.

Calling Every AI Risk a Hoax Collapses Different Problems

“AI safety” is a broad term. It can refer to anything from preventing an ordinary chatbot from producing dangerous instructions to controlling a future autonomous system with capabilities that exceed those of its developers.

The International AI Safety Report 2026, written with contributions from more than 100 experts, separates the debate into several categories. These include malicious use, unreliable or malfunctioning systems, economy-wide effects, and the more speculative possibility that humans could lose control of highly capable AI.

Some of those problems already exist. Generative AI has been used in scams, fraud, impersonation, and cyberattacks. Current models can still hallucinate information, follow malicious instructions hidden in their inputs, or behave unpredictably when they encounter situations outside their training.

Other concerns are forward-looking. Researchers are studying whether increasingly capable models could meaningfully assist with biological threats, automate sophisticated cyber operations, evade oversight, or pursue objectives in ways their operators did not intend. The evidence and probability differ substantially across those scenarios.

Trump’s robot-takeover comment addressed only the most dramatic end of that spectrum. It did not address current problems involving reliability, fraud, cybersecurity, or the deployment of AI in consequential settings.

Two days before the summit call, Anthropic CEO Dario Amodei published an essay titled “The Adolescence of Technology”. Amodei argued that powerful AI could produce major gains in medicine, productivity, and science, while also creating severe risks involving cyberattacks, biological weapons, concentrated power, and labor disruption.

Amodei’s timelines and scenarios are not established facts, nor do all AI researchers agree with them. Forecasting the behavior of systems that do not yet exist inevitably involves uncertainty. Some public warnings also bundle plausible short-term risks with much less certain predictions, which can make careful criticism sound apocalyptic.

But disagreement over probability is not evidence of a hoax. A hoax implies deliberate deception. Current AI failures and malicious uses are measurable, while the more extreme risks remain subjects for research and debate.

The Real Policy Signal Is Speed Over Precaution

Trump’s remarks fit an administration that has consistently placed AI development and infrastructure ahead of precautionary regulation.

Released on July 23, 2025, America’s AI Action Plan listed more than 90 proposed federal actions. Its three main pillars covered innovation, domestic AI infrastructure, and international diplomacy and security. The plan called for reducing regulatory barriers, accelerating data-center construction, expanding energy supplies, and strengthening American leadership abroad.

The plan itself demonstrates why the word “hoax” is too sweeping. One of its pillars is security. The administration’s own policy recognizes that advanced models, computing infrastructure, semiconductor supply chains, and sensitive technologies require protection.

The speakerphone comments came as Trump was also moving to curb state-level AI rules. NBC News reported that the president issued an executive order on September 14 intended to create a national policy framework and limit what the administration regards as burdensome state regulation.

Together, the policies and Trump’s comments point to a clear preference: establish national rules that permit faster development, then challenge state requirements that could slow companies down.

China is central to the argument. Trump suggested that critics of AI could be helping Beijing by delaying American companies. That framing turns nearly every safety proposal into a test of political loyalty rather than a question about technical merit.

Competition and safety do not have to be opposites. Cybersecurity testing, model evaluations, incident reporting, supply-chain controls, and dependable infrastructure can support American competitiveness. An unreliable AI system can be a commercial failure or a national-security weakness, even if it reaches the market before a competitor.

The practical question is not simply whether the United States should develop AI. It is which safeguards improve deployment and which rules add cost without producing a measurable benefit. Calling the entire discussion a hoax prevents that distinction from being made.

Data Centers Are the Tangible Part of the Debate

Trump’s praise for data centers points to a more immediate conflict than hypothetical robot domination. AI development now depends on physical infrastructure: GPUs, networking equipment, cooling systems, electricity generation, substations, and transmission capacity.

A Berkeley Lab analysis of U.S. data-center electricity demand estimated that data centers consumed about 176 terawatt-hours of electricity in 2023, equivalent to 4.4% of total U.S. electricity use. Its projections put consumption between 325 and 580 terawatt-hours by 2028, or roughly 6.7% to 12% of national demand.

That growth can bring construction spending, tax revenue, and investment. It can also require costly grid upgrades and new sources of power. Whether a data center broadly enriches its host community depends on local tax arrangements, utility regulation, energy supply, and how the infrastructure is financed.

Trump’s claim that data centers make states wealthy skips over that distribution problem. The companies building AI infrastructure, the utilities serving it, nearby residents, and taxpayers may not receive the same benefits or absorb the same costs.

This is where the national AI race becomes local. Federal officials can encourage faster permitting, but communities still have to decide how new facilities affect their grids, land, electricity rates, and development plans.

For Nvidia, the expansion is an enormous market opportunity. For governments, it is an industrial and energy strategy. For residents near proposed facilities, it can be a question about bills and infrastructure. Those interests overlap, but they are not identical.

Huang’s Response Shows the Industry’s Balancing Act

Huang’s answer after the call was carefully calibrated. He backed Trump’s urgency, praised the focus on American AI leadership, and supported large-scale infrastructure investment. At the same time, he refused to describe every safety concern as fabricated.

He has also acknowledged that AI could cause substantial workplace disruption. During the summit discussion, Huang estimated that the technology could affect 60% to 70% of jobs by 2030, though he argued that higher productivity could increase overall employment and create new industries. That is Nvidia’s forecast, not a settled economic projection, but it is hardly an argument that AI will have no disruptive effects.

The balance reflects the position of much of the AI industry. Companies want fewer restrictions on construction, chips, model development, and product releases. They also need customers to trust that their systems are secure and reliable.

AI safety is therefore not limited to philosophical discussions about machines surpassing humanity. It includes practical engineering work that businesses already pay for: testing models, preventing data leaks, securing infrastructure, monitoring failures, and limiting abuse.

Huang’s reluctance to adopt the hoax label may have been diplomatic, but it was still notable. The head of a company positioned to gain from faster AI spending endorsed acceleration without pretending that acceleration is risk-free.

Final Thoughts

Trump’s most consequential message was not that robots will fail to conquer the world. It was that broad concern about AI should be viewed as an impediment to American power, potentially encouraged by political opponents or China.

Skepticism toward dramatic AI predictions is reasonable. Forecasts involving superhuman systems, mass unemployment, or loss of control depend on disputed assumptions about technical progress and human behavior. Those claims should be examined rather than accepted because an AI executive made them.

A blanket hoax label is no better. It sweeps documented scams, security failures, unreliable outputs, workplace disruption, and infrastructure costs into the same category as speculative extinction scenarios. It also clashes with an AI strategy that already recognizes national-security and supply-chain risks.

Huang’s answer offered the more defensible position. The United States can build quickly, invest in data centers, and compete with China without treating every safety researcher as an enemy of progress. American leadership will depend not only on how many AI chips it installs, but on whether the resulting systems are reliable enough to deploy without sacrificing security or public trust.