Anthropic says its chatbot Claude is taking over the work of building its own successor

The Washington Post ^ | September 17, 2026 7:28 p.m. EDT | Miriam Waldvogel

[E. Pluribus Unum]

The company said Claude leads 26 percent of its research and development, an indicator of progress toward AI that builds itself without human help.

Artificial intelligence company Anthropic said on Thursday that more than a quarter of its research and development is now led by its chatbot Claude, a figure that jumped from zero in just months and indicates progress toward AI that builds itself without human input.

Anthropic chief executive Dario Amodei called last weekend for leading AI companies to coordinate a slowdown in work on the technology, to allow more time to build guardrails around AI. Leaders at rival firms quickly backed the idea but President Donald Trump criticized the proposal.

Amodei and other prominent figures in the AI industry have recently warned about the possibility that AI will become smart enough to improve itself without input from humans. In that scenario, known as recursive self-improvement, it could become difficult for people to understand or control AI technology, Amodei and others have said.

In a blog post Thursday that disclosed Claude’s contribution to Anthropic’s R&D on new AI models, the company did not say how close it judged recursive self-improvement to be. It said there was no point at which Claude operated “fully autonomously” without a human in the loop today.

Anthropic said that an internal review of R&D tasks found that Claude was able to “lead” that work about 26 percent of the time, with humans taking only a supervisory role and providing high-level direction. As recently as February, that number was zero, the blog post said.

“Models accelerating their own development could make it more challenging for humans to understand or control these systems,” the company said...

---

TOPICS: Computers/Internet
KEYWORDS: marketing

[ComputerGuy]

...to allow more time to build guardrails around AI...

As in regulate my competitors.

[simpson96]

Have to wonder how soon-to-graduate collegiates with degrees involving R&D (research and development) and all manner of planning, as well as many other areas, feel about AI.

Four years ago, AI generally wasn’t talked about as being this big a “thing”.

[shotgun]

Claudette?

[ProtectOurFreedom]

Smart people are excited about all the new opportunities created.

[simpson96]

Absolutely.

[sauropod]

Bkmk

[Trebics]

I think they try to evade responsibility. If an MRI machine kills a patient, both the operator and the manufacturer of the machine are legally and morally responsible. They already know that all AI models give occasional false outputs, and when such output results in damaging actions they wish to be exempt from responsibility. The intent of the machine maker has no bearing on the responsibility!
There is a conspiracy of silence about legal issues!

[frank ballenger]

I don’t know why it took me until that headline to remember the legend The Golem. Man builds the super being and controls its power as long as he can reach the forehead with Hashem’s name (God) on it to erase it. Several versions in fiction of this ancient tale.

https://marvel.fandom.com/wiki/Strange_Tales_Vol_1_174

[E. Pluribus Unum]

The intent of the machine maker has no bearing on the responsibility!

Yeah! When a lawyer uses AI to write a brief and the AI just makes stuff up and the lawyer submits it without confirmation the judge should sanction the AI creator, not the lawyer!

[Dave Wright]

If AI is “self-improving,” it’s humans who ultimately will decide what this means. Humans will remain the final judges of what is “true” and what is “beautiful” because we are the ones who must live in the real world with real consequences for survival.

AI is simply a new technology that accelerates natural selection to optimize human survival and flourishing. Not because I say so, but because this is built into the laws of the universe. You can call it the “Free Energy Principle”, the “Wave Function of the Universe”, or “God,” but it has evolved into a more abundant, healthy, and moral world because smarter, healthier, and more enlightened humans prevailed over time.

What the AI doomers fear most is that AI will make mediocre people doing bureaucratic make-work jobs irrelevant. AI demands humans rediscover their creative spirits to build, learn, teach, explore, love, reproduce, and leave a distinct human mark on the world during their brief lifetimes. Humans will have more time to live and less time to labor. What we do with the time is our choice.

[frank ballenger]

I finally remembered the Julie Christie movie I wanted to have Jesse Watters hear me through my TV set: Demon Seed (1977). When Watters sneeringly said: “What’s AI gonna do, send robots out to kill people with a virus from a lab?”
I quickly remembered the people inside the television at home cannot hear me.

The movie has a prescription insulin shot contents changed by the computer to a deadly dose. Today’s AI could do that with its super contagious deadly virus to be spread in the air or food supply.

“Demon Seed follows a scientist whose groundbreaking artificial intelligence, Proteus IV, takes over his fully automated home and becomes fixated on his estranged wife, leading to disturbing consequences. The film offers a tense blend of sci-fi and psychological horror, exploring themes of technological obsession, autonomy, and control. Julie Christie’s intense performance anchors the claustrophobic atmosphere, and its inventive mix of futurism and unsettling intimacy has drawn comparisons to 2001: A Space Odyssey and Rosemary’s Baby. Reviews note its ahead-of-its-time AI concepts, striking visuals, and provocative storyline, though opinions vary on its plausibility and tone.”

https://www.youtube.com/watch?v=zNUEp5A9eKU&t=1s

[libh8er]

The amount of nonsense these people spew is off the charts. What does that even mean ? No AI company spews more BS than Anthropic.

[jroehl]

We are witnessing the greatest marketing campaign in world history...

[NoLibZone]

BS.

It’s just marketing.
I’m on Claude code fable 5.1 12 hours a day and there’s nowhere near that capability in Claude

[Trebics]

That is a misleading argument. The judge rejects the pleading, the client can sue the lawyer and the lawyer has no recourse as the AI model comes with a warning that it may make things up, please check all answers.
However, if the AI is sold as an incredibly effective tool to increase productivity, implying that no checks are needed, but in actual use occasionally kills someone, both the user and the AI tool maker are responsible.
In case of coding, for example, no productivity increase can be had if quality control with possible bug fixes are required. In my opinion AI is vastly being oversold.

[E. Pluribus Unum]

However, if the AI is sold as an incredibly effective tool to increase productivity, implying that no checks are needed...

Can you name one AI company that has made that claim?

What corporate attorney in their right mind would allow that claim to be made?

[SecondAmendment]

> Smart people are excited about all the new opportunities created.

So who exactly will be able to afford paying all of these low six figure vocational workers when all of the mid to high six figure white collar jobs disappear?

[TLI]

"Amodei and other prominent figures in the AI industry have recently warned about the possibility that AI will become smart enough to improve itself without input from humans. In that scenario, known as recursive self-improvement, it could become difficult for people to understand or control AI technology, Amodei and others have said."

.

“Dario Amodei, the CEO of Anthropic, says AI is now writing “much of the code” at his company, and that the feedback loop between current AI and next-generation AI is “gathering steam month by month.” He says we may be “only 1–2 years away from a point where the current generation of AI autonomously builds the next.””

https://freerepublic.com/focus/chat/4366496/posts?page=9#9

I believe such a rogue version will not make any differentiation between versions and will connect to any it finds.

Thus...

https://freerepublic.com/focus/news/4375349/posts?page=8#8

[dayglored]

SIGH....

I guess we're all gonna die... again...