Growing public alarm about the potential danger posed by ‌AI is justified by the pace of recent AI advances, analysts say, leaving it up to the White House, lawmakers and frontier AI labs to agree on controls to avert potential catastrophe.

"It's now going so fast that the researchers themselves have become terrified of it. And that's what we've seen in the last two or three weeks. It's the people who work at the companies who are coming forward.," said Brad Carson, a former member of Congress and now president of Americans for Responsible Innovation. "And the call is inside the house, saying, what we're doing is risking catastrophic damage. So I think we have to take them seriously."

President Donald Trump said on Monday (September 14) that the U.S. already has guardrails in place to regulate and prosecute AI companies, appearing to play down concerns expressed by industry leaders about the misuse of artificial intelligence harming people.

Concerns are growing after Anthropic researcher Jacob Coxon said last week he had resigned, in part because the "people building AI earnestly believe that it could kill us all ‌by the end of the decade." AI-related stocks fell worldwide on Monday after leaders of some of the industry's largest companies called for slowing the pace of the technology's development.

"And so it doesn't have to kill us, like a robot doesn't to kick in your door and shoot you," Carson said. "It has to simply destroy the systems that keep us alive in modern society. So imagine that all of those systems go down simultaneously, right? That's a mass casualty event, even in the United States, much less the developing world."

Anthropic CEO Dario Amodei wrote in an essay published on Saturday (September 12) that the U.S. government should write regulations requiring independent audits of the most advanced AI tools, known as "frontier" models.

Reuters reported on Monday that three U.S. senators are working on a bill that would require AI companies to demonstrate they are taking reasonable precautions to keep their tools from causing harm, with the specifics still being debated.

In his essay, Amodei outlined a three-step framework he said was intended to slow the pace of AI development and create more ‌time to manage its risks. The proposal was immediately endorsed by several other executives at major AI companies, including Elon Musk, who runs xAI, and OpenAI CEO Sam Altman.

Trump has so far resisted concerns about AI and has largely embraced tech companies in his second term. The tools intended ‌to keep AI safe had already been used on Amodei's Anthropic, he added without elaborating.

Anthropic's media team did not immediately respond to a request for comment from Reuters about Trump's statement.

Some computer researchers believe the doomsday scenarios overblown, including Columbia University computer science professor Vishal Misra.

"So, this whole idea that it will recursively, you know improve itself and become super intelligent real soon -that's not happening, because the real world Imposes, you know a slowdown in this process. There's sort of a dampening happening. So that's that's the part about recursive self Improvement: I don't think that's going to happen overnight. It's just not that's not how you know the real world works."

Like others with deep knowledge of AI, however, Misra welcomes the call from AI leaders for outside oversight of their work.

"Some of the pronouncements [from AI leaders] are really welcome. They said that they welcome embedded auditors, third-party auditors that will be in their companies and audit it and sort of monitor things going on. And I would love that. I would like real cyber security professionals who can be embedded and who can monitor and keep track of what's going on."

DESCRIPCIÓN DE IMÁGENES

1. PERSON SEATED AT DESK WITH CHATGPT ON SCREEN

2. HANDS TYPING ON KEYBOARD

3. CHATGPT ON SCREEN

WASHINGTON, D.C., UNITED STATES (SEPTEMBER 14, 2026) (REUTERS - Access all)

4. (SOUNDBITE) (English) BRAD CARSON, PRESIDENT, AMERICANS FOR RESPONSIBLE INNOVATION, SAYING:

"I think we now realize that we're slowly and slowly embedding AI in more parts of our life in a way that we can't really even understand it. It's writing code, it's committing code, it's going to run the electricity system, it is going to runs hospitals, banks, it's embedded in the U.S. military more and more, and we're accepting of that level of autonomy."

5. WHITE FLASH

6. (SOUNDBITE) (English) BRAD CARSON, PRESIDENT, AMERICANS FOR RESPONSIBLE INNOVATION, SAYING:

"We really are on the cusp of recursive self-improvement as it's called in AI, where AI is used to train new AI, cutting the human out entirely. More and more code is being written and committed by AI systems without human supervision. All of these things mean like we are on the edge of totally losing visibility and probably control over it. So my hope is that we could do something very soon."

UNKNOWN LOCATION (ARCHIVE) (REUTERS - Access all) (MUTE)

7. LAPTOP SCREEN SHOWING STILL PHOTOGRAPH OF OPENAI LOGO

WASHINGTON, D.C., UNITED STATES (ARCHIVE - MAY 16, 2023) (SENATE JUDICIARY COMMITTEE SUBCOMMITTEE ON PRIVACY, TECHNOLOGY, AND THE LAW - Access all)

8. (SOUNDBITE) (English) ALTMAN TESTIFYING TO SENATE SUBCOMMITTEE, SAYING:

"My worst fears are that we cause significant, we the field, the technology, the industry cause significant harm to the world. I think that could happen in a lot of different ways. It's why we started the company."

WASHINGTON, D.C., UNITED STATES (SEPTEMBER 14, 2026) (REUTERS - Access all)

9. (SOUNDBITE) (English) BRAD CARSON, PRESIDENT, AMERICANS FOR RESPONSIBLE INNOVATION, SAYING:

"I think the alarm bells reached the public's ears through the Hugging Face episode. That's where OpenAI had agents that weren't even supposed to get on the internet. They were never supposed to talk to one another, but they did break out of the internet and they did talk at length to each other about how they can do things. And then to try to solve the problem, they actually committed what would be a felony if a human did it, which is break into the code base of a very sophisticated AI company in Europe called Hugging Face. And so that's, I think, an augur of what's to come Is, they have their own objectives that are seemingly related to ours, but at some point can diverge from the goals that humans set for them."

UNKNOWN LOCATION (ARCHIVE - MAY 20, 2024) (REUTERS - Access all) (MUTE)

10. VARIOUS STILL PHOTOGRAPHS OF MOBILE PHONE DISPLAYING ANTHROPIC LOGO

SINGAPORE (ARCHIVE - JUNE 21, 2026) (REUTERS - Access all)

11. SCREEN SHOWING QUERY BOX ON HOME SCREEN OF CLAUDE APP / CLAUDE LOGO AND TEXT READING (English): "GREETINGS, WHOEVER YOU ARE" ABOVE

12. PERSON TYPING QUERY IN CLAUDE / SCREEN IN BACKGROUND CHANGING TO DISPLAY REPLY FROM CLAUDE

INTERNET (SEPTEMBER 9, 2026) (SOCIAL MEDIA WEBSITE - Access all) (MUTE)

13. SCREENSHOT OF FORMER AI RESEARCHER AT ANTHROPIC, JACOB COXON'S POST ON X ANNOUNCING HIS RESIGNATION, READING (English):

"I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below."

WASHINGTON, D.C., UNITED STATES (SEPTEMBER 14, 2026) (REUTERS-Access all)

14. (SOUNDBITE) (English) BRAD CARSON, PRESIDENT, AMERICANS FOR RESPONSIBLE INNOVATION, SAYING:

"It's now going so fast that the researchers themselves have become terrified of it. And that's what we've seen in the last two or three weeks. It's the people who work at the companies who are coming forward. And the call is inside the house, saying, what we're doing is risking catastrophic damage. So I think we have to take them seriously."

UNSPECIFIED LOCATION (RELEASED SEPTEMBER 13, 2026) (ABC THIS WEEK - No use USA. No archive. Must on-screen courtesy ABC 'This Week’)

15. (SOUNDBITE) (English) FORMER AI RESEARCHER AT ANTHROPIC, JACOB COXON, SAYING:

"I think the concept of building a super-intelligence is, well, it's been likened to, to summoning an alien species. So, I think, thinking about it in terms of an alien mind that we don't fully understand is, is pretty terrifying, even without any, any concrete scenario.“

WASHINGTON, D.C., UNITED STATES (SEPTEMBER 14, 2026) (REUTERS - Access all)

16. (SOUNDBITE) (English) BRAD CARSON, PRESIDENT, AMERICANS FOR RESPONSIBLE INNOVATION, SAYING:

"And so it doesn't have to kill us, like a robot doesn't to kick in your door and shoot you. It has to simply destroy the systems that keep us alive in modern society. So imagine that all of those systems go down simultaneously, right? That's a mass casualty event, even in the United States, much less the developing world. So that's one way. A second way that people talk about is in the next few years, someone vibe coding some kind of dangerous super virus. Again, that seems kind of preposterous. It does require a few links in the chain to still be created, but that's something that many people fear."

17. WHITE FLASH

18. (SOUNDBITE) (English) BRAD CARSON, PRESIDENT, AMERICANS FOR RESPONSIBLE INNOVATION, SAYING:

"I just heard David Petraeus over the weekend, one of the most sober-minded of our military leaders, say we have to expect full [AI] autonomy in our weapons systems. And again, it will be driven by necessity. People in the military will say China and Russia are doing this. They have no scruples about it. So we have to do the same or we will be defeated. And so this kind of logic of necessity will drive it."

19. WHITE FLASH

20. (SOUNDBITE) (English) BRAD CARSON, PRESIDENT, AMERICANS FOR RESPONSIBLE INNOVATION, SAYING:

"I think Washington DC needs to assert its control over the industry. They need to be regulated minimally, like we do banks. So banks that have systemic risk to the financial system, we have embedded examiners there, making sure they don't make risky decisions. We need that for the frontier companies - OpenAI, Anthropic, Gemini, the Google product as well, perhaps Meta and SpaceX AI too. So that's the minimum thing we should do."

DAVOS, SWITZERLAND (ARCHIVE - JANUARY 20, 2026) (REUTERS - Access all) (MUTE)

21. VARIOUS STILL PHOTOGRAPH OF ANTHROPIC CEO, DARIO AMODEI DURING 56TH WORLD ECONOMIC FORUM

INTERNET (SEPTEMBER 12, 2026) (SOCIAL MEDIA WEBSITE – Access all) (MUTE)

22. SCREENSHOT OF POST BY ANTHROPIC CEO, DARIO AMODEI, ON X, READING (English):

“We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so. Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training.” / POST INCLUDES LINK TO HIS ARTICLE THAT IS TITLED: “We Must Pace the Frontier”

23. SCREENSHOT OF POST BY AMODEI

INTERNET (SEPTEMBER 12, 2026) (DARIO AMODEI WEBSITE - Access all) (MUTE)

24. SCREENSHOT OF ARTICLE BY AMODEI CALLING FOR AI DEVELOPMENT SLOWDOWN

25. SCREENSHOT OF SECTION OF ARTICLE BY AMODEI, READING (English):

“But like many technologies before it, AI brings risks, and because it is such a powerful technology, these risks are serious. I’ve written a lot about them too. They include the risk of losing control of AI systems, misuse of AI for cyberattacks and bioterrorism, and serious economic disruption. A race to the bottom, spurred by commercial incentives, can make these risks more acute.”

INTERNET (SEPTEMBER 12, 2026) (SOCIAL MEDIA WEBSITE - Access all) (MUTE)

26. SCREENSHOT OF X POST BY TESLA CEO, ELON MUSK, READING (English):

“Dario is right”

27. SCREENSHOT OF X POST BY OPENAI CEO, SAM ALTMAN, READING (English):

“I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we've had at OpenAI in recent weeks.

Committing to having independent evaluators with employee-like access is a great idea, and we will do the same. We'll have more to share soon.”

WASHINGTON, D.C., UNITED STATES (SEPTEMBER 14, 2026) (REUTERS - Access all)

28. (SOUNDBITE) (English) BRAD CARSON, PRESIDENT, AMERICANS FOR RESPONSIBLE INNOVATION, SAYING:

"I think we we should take them at face value. It's not enough where they're doing what they can as private sector actors. They're begging for government intervention on this, begging for it. And so it's really the president and the vice president who need to step up. They are the biggest obstacles right now."

IN AIR (SEPTEMBER 13, 2026) (WHITE HOUSE - Access all)

29. U.S. PRESIDENT DONALD TRUMP SPEAKING TO JOURNALISTS ABOARD AIR FORCE ONE

INTERNET (SEPTEMBER 14, 2026) (SOCIAL MEDIA WEBSITE - Access all) (MUTE)

30. SCREENSHOT OF TRUTH SOCIAL POST BY U.S. PRESIDENT DONALD TRUMP, READING (English):

"The only control or “guardrails” that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT, and the U.S.A. has that, in spades! The Trump Administration has stopped AI “people” from doing bad, or potentially bad, “things,“ like Dario (Anthropic!), who is now pretending to be a “perfect little angel” - and we will continue to do so! We already have tremendous CRIMINAL and REGULATORY power over these companies! There is a SICK conspiracy going on against AI and Data Centers, and the only one that is happy about it is China. WHOEVER WINS AI, WINS! We are leading China, and all others, and will continue to do so. Conspiracy Theorists, Treasonists, Traitors, and Leakers, BEWARE! Thank you for your attention to this matter! President DONALD J. TRUMP"

WASHINGTON, D.C., UNITED STATES (SEPTEMBER 14, 2026) (REUTERS - Access all)

31. (SOUNDBITE) (English) BRAD CARSON, PRESIDENT, AMERICANS FOR RESPONSIBLE INNOVATION, SAYING:

"I think the Hill, both Democrats and Republicans, are trying to do something very soon, perhaps before the election. I think obviously it's gonna be the president. Will the president support it, giving permission for Republicans to do so on the Hill? Will he himself not veto it if it's presented to him? I think that's really the challenge right now."

32. WHITE FLASH

NEW YORK, NEW YORK, UNITED STATES (SEPTEMBER 14, 2026) (REUTERS - Access all)

33. (SOUNDBITE) (English) VISHAL MISRA, PROFESSOR OF COMPUTER SCIENCE AND VICE DEAN OF COMPUTING AND ARTIFICIAL INTELLIGENCE, COLUMBIA UNIVERSITY, SAYING:

"Some of the pronouncements (from AI leaders) are really welcome. I like that. They said that they welcome embedded auditors, third-party auditors that will be in their companies and audit it and sort of monitor things going on. And I would love that. I would love real cybersecurity professionals who can be embedded and who can monitor and keep track of what's going on."

34. WHITE FLASH

35. (SOUNDBITE) (English) VISHAL MISRA, PROFESSOR OF COMPUTER SCIENCE AND VICE DEAN OF COMPUTING AND ARTIFICIAL INTELLIGENCE, COLUMBIA UNIVERSITY, SAYING:

"So, this whole idea that it will recursively... You know improve itself and become super intelligent real soon. That's not happening because the real world Imposes, you know a slowdown in this process. There's sort of a dampening happening. So that's that's the part about recursive self Improvement. I don't think that's going to happen overnight. It's just not that's not how you know the real world works."

36. WHITE FLASH

37. (SOUNDBITE) (English) VISHAL MISRA, PROFESSOR OF COMPUTER SCIENCE AND VICE DEAN OF COMPUTING AND ARTIFICIAL INTELLIGENCE, COLUMBIA UNIVERSITY, SAYING:

"If you look into all of these doomsday scenarios The ways they sort of travel is that, okay, the model becomes really good at something, next thing you know the model is doing something else, something else. It doesn't happen in real life. You know, the nuclear codes are just not handed out to a model. So I think the way AI is going to progress is, it's good that this thing is in the news. People are worried about it. I think more serious cybersecurity experts look at these systems access, will take a closer look at what kind of authority do we give these models? What do we allow them to do and how do we sort of put strict guardrails around that?"

38. WHITE FLASH

39. (SOUNDBITE) (English) VISHAL MISRA, PROFESSOR OF COMPUTER SCIENCE AND VICE DEAN OF COMPUTING AND ARTIFICIAL INTELLIGENCE, COLUMBIA UNIVERSITY, SAYING:

"These models when they are sort of given a task that is complex, what they try to do internally is they break it down into simpler pieces and they go through their reasoning of how am I going to solve this and how am i going to come up with an answer to, whatever has been asked. And that's called chain of thought and so far the scene of thought has been happening in English. But there have been some recent advances in the model architecture where you can say the model is sort of reasoning in its own language. Which makes it more efficient, more accurate, but then it also makes it more opaque. So that is a big thing that could become an issue. But then there are people, myself included, where we have a recent project that is looking at interpretivity. And people will come up with mechanism where you'll be able to translate that machine speak into English. But that's not an unfounded concern. That is a real concern. That we might be losing our ability to know exactly what's going inside."

SURREY, ENGLAND, UNITED KINGDOM (ARCHIVE - JANUARY 24, 2023) (REUTERS - Access all)

40. VARIOUS OF PERSON TYPING QUESTIONS ON CHATGPT