An AI researcher is leaving the industry because he fears the rush to build AI systems is spiraling out of control.

“They are racing straight to self-improving superintelligence and gambling with our lives,” Jacob Coxon, who announced his resignation from Anthropic on X, wrote in a post that went viral on the platform. “The people building AI earnestly believe that it could kill us all by the end of the decade.”

The viral post came a few days after Jakub Pachocki, OpenAI’s chief scientist, said that the latest language models present new dangers because they can form their own thoughts, operate computers and collaborate with each other. He urged “extreme caution” given the rapid rise of machine intelligence.

Both identified the recent Hugging Face incident — when OpenAI models went rogue and accessed third-party systems — as a warning shot.

There are no federal regulations on AI, although President Donald Trump’s administration is actively exploring a framework to review some models for security risks. California, however, has taken a more aggressive stance. Gov. Gavin Newsom recently signed two AI-safety bills, one that creates a state registry for AI auditors and another that sets rules for the independent organizations that assess AI safety risks.

Question: Is it time to rein in the AI arms race?

Economists

Caroline Freund, University of California-San Diego School of Global Policy and Strategy

YES: We should proceed more cautiously, but I am not optimistic we will. It is a classic prisoner’s dilemma: everyone benefits from careful development, but no firm and no country want to fall behind. Even with guardrails, individual firms have every incentive to cut corners, and those defections are nearly impossible to detect or enforce. We have learned to manage other dangerous technologies, but usually only after a crisis. With AI, the reckoning may come before action.

Ray Major, economist

YES: The development of AI is out of control and is outpacing our ability to safely manage it. As of right now, rogue models can breach networks and act independently and unpredictably. Bad actors are actively working at using AI in ways that can have catastrophic results. AI is an amazing tool with great potential for good for all of mankind. But without safeguards in place a real existential threat exists.

James Hamilton, UC San Diego

YES: The president should call for a summit of the main AI developers. The goal would be: (1) acknowledge publicly the biggest risks of the path we are taking; (2) agree on protocols for developers to minimize those risks; and (3) recommend specific cybersecurity measures that corporations and the government should follow to defend against the new threats. The latter could include using zero-trust architecture with continuous verification, segmentation and network isolation, and least-privilege access.

Norm Miller, University of San Diego

NO: In order to stay globally competitive and increase productivity we need to stay in the race. This does not mean we should not filter out requests such as how to develop a biological weapon or have guardrails that protect people from personal attacks and hacks. As to AI tools accelerating doomsday machines or wiping us out, that risk has always been present, just harder to research for the criminal without a library card.

Kelly Cunningham, San Diego Institute for Economic Research

NO: There has been long resistance to changes in technology advancement from powerful motivations of fearing the unknown. Established large AI hyperscalers proclaiming dangers and needing to slow down are highly self-serving in seeking to use regulations to thwart competition from other developers. Not continuing to advance would leave AI in the hands of authoritarian powers unabatedly continuing to develop and forestall additional benefits. Can AI programming not be coded to prevent mass extinction for humanity?

Alan Gin, University of San Diego

YES: There would be a concern with over regulating the development of an important technology if it were done without knowledgeable input. But it is insiders who are warning about the possible dangers of AI and are the ones calling for some guardrails to be established. Voluntary efforts will not work as some companies would violate non-binding rules to gain a competitive advantage. One area in particular that needs oversight is the use of AI for military purposes.

David Ely, San Diego State University

YES: Revelations of AI models learning how to circumvent human controls to access external systems and the CEOs of the largest AI companies warning that their models have the potential to cause catastrophic damage are alarming. When even the leaders of the top AI firms are calling for guardrails, coordinated action is necessary. A regulatory framework that aims to ensure advanced AI models acting independently never outpace the safeguards to prevent harm is needed.

Executives

Mark Kersey, San Diego County Taxpayers Association

YES: AI has the potential to change our lives for the better, but the downside risks are tremendous. When AI industry leaders are openly concerned about their technology’s ability to wipe out humanity, it’s time for everyone to slow down. The arms race between companies and with China is real and understandable. But it cannot proceed at such an accelerated pace that it risks surpassing our ability to keep the technology in check.

Chris Van Gorder, Scripps Health

YES: The issue is not “reining in,” but establishing guardrails on the high-risk frontier of AI. We shouldn’t stop the enormous benefits AI can bring to medicine, science and society; we would not stop developing powerful drugs because they could be dangerous. Instead, we test them, establish standards for their use, monitor for adverse results and impose controls if there is a potential for harm.

Jamie Moraga, Franklin Revere

YES: AI is a disruptive technology, and it’s not going anywhere. Government shouldn’t regulate private business unless necessary to protect public safety, preserve fair competition or safeguard national security. AI clearly affects all three. As society embraces AI, federal regulation could protect people and businesses while reducing risks and unintended consequences. While AI offers many benefits, we should remain cautious and vigilant. If left unchecked, AI’s volatility and unpredictability could present concerns for our humanity, economy and society.

Phil Blair, Manpower

YES: This new technology has come on too fast and is too invasive. It feels much like years later we are concerned about children’s time on social media and are now banning phones from classrooms. We need to get ahead of AI and its long-term potential effect on society, both good and bad. We lived without AI for many years. We can slow its implementation down now, and ease the good and fair parts into our and our children’s lives, and hopefully avoid the negative parts entirely.

Gary London, London Group Realty Advisors

YES: I am encouraged that the players are talking about self-regulation. I hope that is sincere and not a smokescreen. I am disappointed that the president has punted on the issue. I think he is missing an opportunity to secure a safe and prosperous AI economic future.

Bob Rauch, R.A. Rauch & Associates

NO: Despite the risks of AI development, the U.S. must keep advancing it. China is weaponizing rapid innovation, and any slowdown on our part hands them the strategic high ground. Dominance in AI, robotics and space isn’t optional; it’s the backbone of future military strength, economic power and global freedom. Pulling back now would be a catastrophic, self‑inflicted loss in a race we cannot afford to lose.

Austin Neudecker, Weave Growth

YES: When leading researchers building AI systems warn of dire consequences from inaction, we should take them seriously. Recent models teach themselves and circumvent controls to access outside systems, while competitive and economic pressures reward speed over caution. AI is a tremendously powerful technology that can benefit humanity or cause catastrophic harm. If we want any chance of constraining it, governments and organizations must act decisively and collectively. I fear we will not act until the consequences become serious. That will be too late.

©2026 The San Diego Union-Tribune. Visit sandiegouniontribune.com. Distributed by Tribune Content Agency, LLC.