Is Recursive Self-Improvement A Current AI Capability Or A Future Concern?
AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: Is Recursive Self-Improvement A Current AI Capability Or A Future Concern? on ThorstenMeyerAI.com

Before you orderOffer from Amazon

Get hardware and tech essentials delivered free with Prime

  • Fast, free delivery on millions of items
  • Prime Video, Amazon Music and more included
  • Member-only deals all year
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

TL;DR

Anthropic says Claude leads 26% of its AI research and development work, while no measured subset runs fully autonomously. METR estimates Anthropic is improving models on well-specified tasks at around twice its earlier rate, with high uncertainty and much of the acceleration likely tied to AI assistance. A congressional bill would pause some AI development, but it has not advanced.

AI is helping researchers build AI faster, but evidence that a system can independently design and develop its own successor has not been documented, according to Anthropic and independent evaluator METR. The distinction is at the center of renewed debate after podcaster Dwarkesh Patel published an essay on November 13 arguing that AI is already accelerating AI development, while a congressional bill proposing limits on that process remains stalled in committee.

Anthropic reported that, as of August 2026, Claude leads 26% of the company’s AI research and development work. The company said more than 90% of that work involved AI at or above a level where it collaborates with people, but that Claude was not operating fully autonomously in any measured subset. Anthropic defines the more complete form of recursive self-improvement as an AI system that can autonomously design and develop its own successor.

METR, a nonprofit evaluator, examined Anthropic’s internal data from August 12 to October 29. In a November 5 report, it estimated that Anthropic was improving model capabilities on well-specified tasks at around twice the rate recorded from March 2024 to October 2025. METR stressed that the estimate carries high uncertainty, saying the actual acceleration could be closer to 1.3 times or three times. It judged that AI assistance was likely responsible for a majority of the speed-up, while saying AI had not yet made highly significant, fully autonomous contributions to the most important parts of model training.

The congressional proposal is the Ban Artificial Superintelligence Act of 2026, introduced by Sen. Bernie Sanders as S.5493 on September 23 and Rep. Greg Casar as H.R.10538 the following day. It would pause training, modifying and fine-tuning large AI systems until a new Department of Artificial Intelligence sets rules. The bill also targets systems capable of automating or greatly accelerating AI research and development. Both versions remain referred to committee; the House bill has 12 Democratic cosponsors, according to the source material.

At a glance
analysisWhen: As of November 16, 2026; bill introduce…
The developmentViral discussion of recursive self-improvement has renewed attention to evidence that AI is helping with AI research and to a congressional bill that would restrict parts of that work.
Crypto market snapshot
Fear & Greed Index
10/100 — Extreme Fear
Bitcoin BTC$64,883▲ 0.7%
Ethereum ETH$1,818▲ 1.3%
Tether USDT$0.9995▲ 0.0%
BNB BNB$593.85▲ 1.1%
USDC USDC$0.9997▲ 0.0%
XRP XRP$0.9654▲ 1.8%
Solana SOL$71.9▲ 1.7%
TRON TRX$0.3037▼ 1.1%
Live data · CoinGecko · alternative.me (24h change)

Has it started? · explainer

AI already does much of the work of building AI. Not alone.

How far AI research and development is automated at one lab, on a scale from AL0 (no AI) to AL5 (no human in the loop), and where the proposed ban would draw its line.

Anthropic’s AI R&D work, August 2026

Automation levels from Epoch AI’s scale, as measured by Anthropic

AI collaborates or moreabove 90%
AL3 and up
AI leads, a human supervises26%
AL4
Fully autonomousno measured part
AL5 · the closed loop

Anthropic: “We are not there yet.”

The bill’s line: “the capacity to automate or greatly accelerate” AI R&DS.5493 / H.R.10538 list it as a “superintelligence precursor”. How much is “greatly” is not defined.
  1. AI collaborates on most of the workAbove 90% of the work is at or above “AI collaborates”, where AI does large chunks under close human direction.
  2. AI leads about a quarterIn 26%, Claude completes most of a task from a high-level prompt while a human supervises.
  3. No closed loopNo measured part runs fully autonomously. That is the step “recursive self-improvement” usually means.
  4. Where would a ban start?The bill names “greatly accelerate” as a warning sign, so its reach depends on a measurement nobody yet makes in a standard way.

Sources: Anthropic, “Measurements for understanding the pace of AI development inside frontier labs” and “When AI builds itself” (figures as of August 2026, Anthropic’s own measurements); Ban Artificial Superintelligence Act of 2026, S.5493 and H.R.10538, sec. 3(6)(A). Bars show shares of Anthropic’s AI R&D work.

Graphic: Powered by Thorsten Meyer AI

Why AI-Assisted Research Matters

The distinction between AI assistance and autonomous self-improvement concerns different levels of capability and potentially different policy responses. Current reports describe AI helping researchers with parts of development while people continue to set priorities, supervise work and make decisions. They do not establish that AI systems can independently run the research cycle or create successor models without human direction.

AI assistance could affect the pace of research even when the process remains human-led. METR’s estimate suggests that AI use may be contributing to a higher pace of progress on tasks that can be clearly specified and evaluated. The estimate has substantial uncertainty, applies to a limited set of tasks and draws on internal company data. The evidence does not establish that the same rate applies across AI research or will continue.

If enacted, the proposed law could affect companies and researchers because its scope extends beyond systems explicitly building their own successors. The text also covers systems able to automate or greatly accelerate AI research and development. For now, the proposal has not changed rules for businesses or users; it has not advanced beyond committee referral.

Products Worth Considering

Apple MacBook Air 15-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, Liquid Retina Display, and Touch ID (Midnight)

Apple MacBook Air 15-inch Laptop with M5 Chip, 16GB Memory, 512GB SSD, Liquid Retina Display, and Touch ID (Midnight)

Apple MacBook Pro 16.2-inch with M5 Pro, 18-core CPU, 20-core GPU, 24GB Memory, 1TB SSD, Wi-Fi 7, Space Black

Apple MacBook Pro 16.2-inch with M5 Pro, 18-core CPU, 20-core GPU, 24GB Memory, 1TB SSD, Wi-Fi 7, Space Black

As an affiliate, we earn on qualifying purchases.

From Research Assistance to RSI

Recursive self-improvement, or RSI, generally describes a feedback loop in which an AI system helps develop a more capable system, which in turn helps with later development. Anthropic’s definition focuses on the point at which an AI can fully autonomously design and develop its successor. The term is also being used more broadly in public debate to describe AI tools helping human researchers with coding, experiments or other development tasks.

Those meanings should not be treated as interchangeable. Anthropic uses an automation scale ranging from no AI involvement to fully autonomous operation. On that scale, collaboration means AI works with people, while the “leads” level means it can complete most of a task from a high-level prompt under human supervision. Anthropic’s reported 26% figure refers to this supervised “leads” category—not autonomous model development.

Patel’s November 13 essay, titled “The avalanche that hit math is coming for AI development,” draws on METR’s assessment and figures published by Anthropic and OpenAI, then makes its own forecasts about what could follow. The essay is commentary rather than new experimental evidence. The debate has gained political attention after Democrats won both chambers of Congress on November 3, according to the source material, though the current bill status remains unchanged.

“We are not there yet, and recursive self-improvement is not inevitable. But it could come sooner than most institutions are prepared for.”

— Anthropic, in its essay “When AI builds itself”

Limits of the Current Evidence

Full autonomous self-improvement has not been demonstrated in the evidence cited here, but that finding is limited to what was reported and assessed; it does not establish what may happen in the future. METR says its estimate of Anthropic’s acceleration is highly uncertain and notes that Anthropic redacted a majority of the evidence METR found most informative. The report relies on internal company data, so readers cannot independently inspect all the underlying material.

OpenAI has also described progress toward automated research. According to The Decoder, which reviewed an OpenAI post that the source publisher could not download, the company said it had reached its “automated research intern” goal by its own measurements and aims for a full automated AI researcher by March 2028. The reported figures include success on some short tasks without intervention, but longer tasks often still needed a human step. The Decoder said OpenAI provided no independent review of those figures. The scope and comparability of the companies’ measurements are not established by the material available here.

The House and Senate bills’ committee referrals do not show whether lawmakers will advance them, amend them or take no further action. A separate proposal from Rep. Ro Khanna, called the Human Control Over AI Act, was described in a September CNBC report as a summary; the source material says no bill text was found. Its specific provisions and status therefore remain unclear.

Evidence and Legislative Steps Ahead

The next concrete policy milestone would be action in the committees handling S.5493 and H.R.10538. As of November 16, the bills had not moved beyond referral, and no timetable for hearings, votes or further consideration was included in the source material. Their language and scope could change if lawmakers take them up.

On the research side, further independent evaluation could help clarify how much AI assistance changes the pace of development, which tasks are included in company estimates and how much human oversight remains. OpenAI’s stated goal is a full automated AI researcher by March 2028, but that is a company target, not a demonstrated capability or a guaranteed timeline. For now, the evidence supports a narrower conclusion: AI is contributing to research work, while the fully autonomous version of recursive self-improvement remains unverified.

Key Questions

Has recursive self-improvement already started?

That depends on how the term is defined. AI is assisting and leading some AI research tasks under human supervision, according to Anthropic. A system independently designing and developing its own successor has not been documented in the evidence cited here.

What does METR’s estimate of around twice the rate mean?

METR estimated that Anthropic improved model capabilities on well-specified tasks at around twice the rate recorded between March 2024 and October 2025. The evaluator described the estimate as highly uncertain, with a possible range closer to 1.3 times or three times.

Would the proposed federal bill ban all AI research?

The Ban Artificial Superintelligence Act proposes pausing training, modification and fine-tuning of large AI systems until a new federal department sets rules. It also addresses systems able to automate or greatly accelerate AI research and development. It is a proposal, not current law, and remains in committee.

When does OpenAI expect a full automated AI researcher?

According to The Decoder’s account of an OpenAI post, the company has set March 2028 as its goal for a full automated AI researcher. That is a target, not confirmation that the capability will be achieved by then.

Source: ThorstenMeyerAI.com

Nothing in this article is financial or investment advice. Cryptocurrency and precious-metal investments carry significant risk — do your own research and consider a licensed advisor.
You May Also Like

Running Frontier AI On A Mac Studio: What Does It Really Take?

Analyzing Apple’s new Mac Studio with 512GB memory for local frontier-scale AI model inference, including capabilities, limitations, and what it means for users.

Britain Pays More Than The US To Check AI Before Launch

The UK Budget adds £120 million for AISI over four years. Annualized, UK funding is about eight to nine times the US center’s current budget.

What Strong Authorization Means For AI Agent Security

A proposed proxy would require approval for each high-risk AI agent action and log grants and denials. Its security and latency benefits remain untested.

Grok Wants To Keep ‘Nudifying’ Images As It Sues Minnesota

A Baptist News Global headline reports that xAI is suing Minnesota over restrictions tied to Grok and sexualized images; key case details remain unavailable.