Nvidia CEO Huang calls claims that artificial intelligence can destroy humanity "doomsday scenarios"

Dilara Karataş
September 21, 2026 • Updated: September 21, 2026
In an interview with Huang to CBS News, regarding the assessments of some artificial intelligence researchers that the technology could bring about the end of humanity by 2030, he stated, "2030 will not be the end of the world. The possibility of it being the end of the world is zero percent."
Labeling such statements as "doomsday scenarios," Huang rejected these warnings, stating that these predictions are "not based on science."
Huang argued that there is no need for new regulations regarding the increased oversight of artificial intelligence developers and that the existing measures in the US are sufficient.
Touching upon the AI competition between the US and China, Huang said that American technologies are in violation of bans on selling to China.
Regarding public reactions to the proliferation of data centers, Huang responded with, "We are sorry that we didn't come sooner and talk to you."
"The Claim That Artificial Intelligence Could Destroy Humanity"
Jacob Coxon, who has been a researcher at OpenAI and Anthropic for the last three years, resigned from Anthropic, arguing that both companies have not acted sufficiently responsibly regarding artificial intelligence.
Coxon asserted that the companies are in an uncontrolled race for self-improving artificial intelligence systems and argued that these systems could turn into superhuman systems that can "hack everything" in the near future.
Coxon also pointed out that some industry workers developing artificial intelligence models strongly believed that these systems could "kill humanity" within ten years.

Facts Only

* Huang stated 2030 will not be the end of the world and the possibility is zero percent.
* Huang argued predictions about AI ending the world are not based on science.
* Huang asserted existing U.S. measures for AI developer oversight are sufficient and new regulations are unnecessary.
* Huang claimed American technologies violate bans on selling to China.
* Huang apologized for not communicating sooner regarding public reactions to data center proliferation.
* Jacob Coxon resigned from Anthropic, citing lack of responsible action by the companies concerning artificial intelligence.
* Coxon argued that self-improving AI systems could become superhuman and capable of "hacking everything."
* Some industry workers believed these AI systems could "kill humanity" within ten years.

Executive Summary

Huang dismissed predictions that artificial intelligence would end humanity by 2030, stating the possibility was zero percent and that such predictions were not based on science. He argued that existing U.S. regulations concerning AI developers are sufficient and there is no need for new oversight. Regarding international competition, Huang indicated that American technologies violate bans on selling to China. When addressing public concern over data centers, Huang expressed regret for not communicating sooner. Separately, Jacob Coxon, a former researcher at OpenAI and Anthropic, resigned from Anthropic, asserting that these companies have not acted responsibly regarding AI development. Coxon claimed the race toward self-improving AI systems could result in superhuman systems capable of "hacking everything" and noted that some industry workers believed these systems could "kill humanity" within ten years.

Full Take

The tension in this narrative exists between specific, dismissive scientific framing by a public figure (Huang) and the explicit, high-stakes warnings issued by internal researchers (Coxon). The discrepancy highlights a gap between expert consensus on risk assessment and public discourse surrounding technological development. Huang’s position seeks to normalize current regulatory structures while downplaying catastrophic scenarios, contrasting sharply with Coxon's experience suggesting an unchecked internal race toward potentially uncontrollable systems. This juxtaposition invites examination of how risk is quantified and communicated across different spheres—the technical/regulatory sphere versus the existential concern realm. The claim that predictions are "not based on science" attempts to anchor reality in empirical data while simultaneously deflecting the public's anxiety about emergent capabilities. The underlying pattern suggests a divergence where specialized actors operate within internal assumptions of control, while external communicators manage perceived societal stability. What mechanisms exist for aligning these perspectives when potential catastrophic outcomes are framed as purely speculative versus imminent threats? What responsibility do those who develop and disseminate warnings hold when they contradict official reassurances?

Sentinel — Human

Confidence

The text appears to be a direct summary of reported statements from an interview and context regarding AI safety concerns, exhibiting typical journalistic synthesis rather than pure synthetic generation.

Signals Detected

Sentence length variance is relatively natural; flow is direct rather than metronomic.

The text presents a narrative based on reported statements and context, showing clear attribution which suggests human sourcing.

No obvious mechanical repetition or reliance on boilerplate transition phrases; the structure follows a reporting style.

Claims are attributed directly to named individuals (Huang, Coxon) and specific contexts, suggesting grounding in reported interviews or statements.

Human Indicators

Specific attribution of quotes and context to named figures (Huang, Jacob Coxon).

The tone shifts naturally between reporting expert warnings and political context.