Is it possible for artificial intelligence to wipe out humanity before 2030? Jensen Huang’s answer is an emphatic “0%.” As debates over AI losing control intensify, he does not deny the existence of safety risks, but he pushes back against “doomsday narratives” lacking factual and scientific grounding that could slow down the entire industry.
NVIDIA (NVDA.O) co-founder and CEO Jensen Huang has once again publicly rejected AI “doomsday” theories. In an interview aired on CBS News on September 20, he flatly dismissed claims that AI would lead to human extinction before the end of this decade, calling such predictions devoid of scientific and factual basis. However, he stressed that AI companies must dedicate sufficient resources to address safety concerns and take responsibility for any real harm caused by their products.
Huang’s remarks come at a time when safety debates within the U.S. AI industry have grown noticeably louder. Over the past several weeks, numerous AI researchers and corporate leaders have publicly discussed the risks of frontier models spiraling out of control, with some advocating for a slowdown in capability advancement and stricter safety protocols. Huang does not dispute that these risks require attention, but he objects to expanding the discussion into unsupported predictions of human extinction and rejects using such fears as a justification to decelerate industry progress.
No Doomsday in 2030
When a CBS reporter asked whether AI could cause humanity’s disappearance within the next decade, Huang directly called the notion “completely wrong.” He stated that 2030 will not mark the end of the world and put the probability at “0%.” He went on to draw a clear line between legitimate AI safety concerns and the “doomsday narrative” surrounding the technology. In Huang’s view, worrying about safety is not misplaced, but certain catastrophic predictions are “not built on facts or science.” He argued that as AI transitions from laboratory research to large-scale products and services, developers need to channel more researchers and computing resources into safety work rather than relying on extreme risk narratives to drive regulation.
A direct backdrop to the recent debate is former Anthropic researcher Jacob Coxon’s public statement that some AI developers genuinely fear AI could cause catastrophic outcomes before the decade’s end. Anthropic CEO Dario Amodei subsequently proposed slowing the pace of frontier AI capability development and called for more systematic safety arrangements between companies and the U.S. government. OpenAI CEO Sam Altman and other industry figures have also publicly supported heightened attention to risk.
Huang is taking a different path. Over the past week, he has repeatedly emphasized that AI safety is fundamentally an engineering problem, and that developers have a duty to complete necessary testing and safeguards before deploying products. In the CBS interview, he said the industry should push AI forward “as fast as possible,” but speed cannot come at the expense of safety. If a company releases an unsafe product that causes harm to people, existing laws are already sufficient to hold it accountable.
This also forms the core of Huang’s regulatory stance. He is not arguing that AI needs no oversight; rather, he believes existing legal frameworks—such as product liability, cybersecurity, and unauthorized access rules—should be enforced first. Huang stated that the U.S. “doesn’t need more regulation,” but rather application of the rules already on the books.
For some frontier AI companies calling for greater government intervention, Huang’s criticism is more pointed. In the CBS interview, he said that companies warning about severe risks from their own technology while simultaneously demanding new regulatory arrangements make him question their motives. He also suggested that some industry proposals might actually be aimed at reducing existing legal liabilities. That said, he provided no evidence that these companies harbor specific political or commercial agendas, and he acknowledged he does not know their exact intentions.
Recent Share Changes Not Entirely Active Selling
While Huang has been vocal on AI policy, his personal shareholding changes have also drawn market attention. A Form 4 filing disclosed by the U.S. Securities and Exchange Commission (SEC) on September 18 shows that some of Huang’s recently reported reductions in NVIDIA stock were not completed through open-market active sales. According to the filing, on September 16, NVIDIA withheld 45,728 ordinary shares from Huang at a reported price of $212.17 per share. The SEC document explicitly states that these shares were withheld by the company to cover taxes due from the vesting of previously disclosed restricted stock units. As such, this transaction is a tax withholding event, not an active sale by Huang on the secondary market.
On September 17, an additional 438,000 shares were transferred out of Huang-related trusts as gifts, with a reported price of zero. Of these, 292,000 shares went to the Jen-Hsun & Lori Huang Foundation, while 146,000 shares were moved into a donor-advised fund. The SEC filing clearly categorizes both transactions as gifts with no consideration. Therefore, judging Huang’s holdings purely by the reduction in share count would inaccurately suggest a concentrated cash-out. The latest Form 4 includes both tax withholdings and charitable transfers, both of which are materially different from active open-market selling. The filing shows that after the September 16 transactions, Huang still directly holds approximately 70.1 million NVIDIA shares, with additional substantial holdings held indirectly through various trusts and limited liability companies.