AIThis post was created with the assistance of artificial intelligence (AI).

🔍 Read the full analysis: A Ban On AI Improvement: Would It Cover Recursive Self-Improvement? on ThorstenMeyerAI.com

Buying for a business?Offer from Amazon

Get business pricing on monitors, keyboards and dev gear

  • Business-only prices and quantity discounts
  • Tax-exempt purchasing
  • Multiple users, one account, clear invoices
As an affiliate, we earn on qualifying purchases.

TL;DR

A proposed bill in Congress would pause training and modification of large AI systems, explicitly including recursive self-improvement, until a new federal agency sets rules. Current evidence describes AI helping researchers and leading parts of the work, not autonomously building its own successor; the bills have not advanced beyond committee referral.

A viral debate over whether AI is already helping build more capable AI has brought renewed attention to the Ban Artificial Superintelligence Act of 2026, a bill that would pause specified AI development activities, including recursive self-improvement. The proposal remains at the committee stage, while available evidence describes AI assisting human researchers—not independently designing and developing its own successor.

The bill was introduced by Sen. Bernie Sanders as S.5493 on September 23 and by Rep. Greg Casar as H.R.10538 the following day. Its text would pause training, modification and fine-tuning of large AI systems until a proposed Department of Artificial Intelligence establishes rules. It also covers systems with the capacity to automate or greatly accelerate AI research and development. Both bills have been referred to committees and have not advanced. The House version has 12 Democratic cosponsors, according to the status information cited in the source material.

The debate intensified after podcaster Dwarkesh Patel published an essay on November 13 arguing that AI is already speeding up AI development and forecasting much faster progress. His account draws on evaluations and company-reported figures, but the forecast is his interpretation, not new independent evidence that a system has built its own successor.

Anthropic reported that, as of August 2026, Claude led 26% of its AI research and development work. The company also said more than 90% of that work involved AI at or above a collaboration level, while no measured subset was fully autonomous. The figures are company-reported; they describe the degree of AI involvement in work, not a self-directed cycle in which a system independently sets goals and develops its replacement.

At a glance
reportWhen: Bills introduced September 2026; debate…
The developmentA viral debate over whether AI is already improving itself has renewed attention to a congressional bill that would restrict AI development, including recursive self-improvement.

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

What the Proposed Pause Would Cover

The proposal could reach beyond a narrow prohibition on a machine autonomously designing its successor. Its text would pause training and modification of large systems and cover systems capable of automating or greatly accelerating AI research and development. That broader language makes the practical scope of any future restrictions consequential for AI labs, researchers and organizations using advanced models.

For readers, the distinction is between AI-assisted research and a fully autonomous improvement loop. AI tools can help with experiments or complete portions of research while people choose objectives, review results and decide whether to train or deploy a model. A ban written to cover acceleration of research could apply to some activities well before the more dramatic scenario of a system building its own successor is demonstrated. The bill’s status means it creates no current restrictions for businesses or users.

Products Worth Considering

Bose QuietComfort Headphones - Wireless Bluetooth Over-Ear Noise Cancelling Headphones, Petal Pink Limited Edition

Bose QuietComfort Headphones – Wireless Bluetooth Over-Ear Noise Cancelling Headphones, Petal Pink Limited Edition

Amazon Kindle Paperwhite Signature Edition (2024), 32 GB, Metallic Black

Amazon Kindle Paperwhite Signature Edition (2024), 32 GB, Metallic Black

As an affiliate, we earn on qualifying purchases.

Evidence Behind the Self-Improvement Debate

Recursive self-improvement is often used to describe a cycle in which an AI system helps create a better system, which then helps create the next one. Anthropic defines the far end of that spectrum as an AI capable of fully autonomously designing and developing its own successor. The company says that point has not been reached. A wide range of human involvement lies between no AI assistance and full autonomy, so claims that self-improvement has “started” depend on how the phrase is defined.

In a November 5 assessment based on internal Anthropic data from August 12 to October 29, the independent evaluator METR estimated that Anthropic was improving model capabilities on well-specified tasks at around twice the rate seen between March 2024 and October 2025. METR described the estimate as highly uncertain, saying the true acceleration could be closer to 1.3 times or three times. It judged that AI assistance likely accounted for most of the faster pace, while saying AI was not yet making highly significant, fully autonomous contributions to the most important parts of model training. METR also said Anthropic had redacted much of the evidence it considered most informative.

OpenAI has separately published internal measures of research automation. The Decoder, which reviewed the company’s post after its website blocked the original download for the source publication, reported that OpenAI said it had reached its “automated research intern” goal and aims for a fully automated AI researcher by March 2028. The reported results included high success on tasks shorter than 15 minutes, but more than half of successful tasks lasting four to eight human hours required at least one human step. These are company measurements, with no independent review identified in The Decoder’s account.

““We are not there yet, and recursive self-improvement is not inevitable.””

— Anthropic, in its essay “When AI Builds Itself”

Where the Evidence and Bill Leave Gaps

There is no documented evidence in the source material of an AI system autonomously designing and developing its own successor. But the boundaries of the proposed bill are not settled in practice: its language covers systems able to automate or greatly accelerate AI research, and the source material does not establish how regulators would define those thresholds or apply them to specific tools and workflows.

The evidence also has limitations. METR’s acceleration estimate carries substantial uncertainty and relies on internal company data, much of which it said was redacted. OpenAI’s figures are self-reported, and the cited account found no independent review. The status of Rep. Ro Khanna’s separate “Human Control Over AI Act” is less clear: CNBC saw a summary in September, but the source material reports that no bill text was found. None of these details establishes how Congress would resolve disagreements over what counts as recursive self-improvement.

Committee Action and Definitions Ahead

The immediate next development is whether the House and Senate committees receiving the bills take them up. As of the latest status information in the source material, neither has moved beyond referral, and the proposals impose no current change on AI companies or users. Any legislative progress would bring further attention to what counts as a large AI system, research acceleration and recursive self-improvement.

Further evaluations may clarify how much AI contributes to model development and which tasks remain under human direction. For now, the central distinction remains: AI is assisting and, in some reported work, leading parts of research, while autonomous development of a successor has not been shown in the cited evidence. The timing and content of any committee action, and whether a separate proposal receives full bill text, remain unknown.

Key Questions

Has recursive self-improvement been demonstrated?

The cited evidence does not show an AI system autonomously designing and developing its own successor. It does describe AI assisting researchers and leading portions of some AI research and development work.

What would the Ban Artificial Superintelligence Act restrict?

The proposed bill would pause training, modification and fine-tuning of large AI systems until a proposed Department of Artificial Intelligence sets rules. It also covers systems with the capacity to automate or greatly accelerate AI research and development.

Is the bill law or in effect now?

No. The House and Senate bills have been referred to committees and have not advanced, so they do not currently impose restrictions on companies or users.

How reliable is the estimate that AI sped up Anthropic’s work?

METR estimated that Anthropic’s progress on well-specified tasks was around twice its earlier rate, but called the estimate highly uncertain. Its assessment used internal company data, and METR said much of the evidence it found most informative was redacted.

Source: ThorstenMeyerAI.com

You May Also Like

Why Is Anthropic Spending $6 Billion On Decart? Exploring The Future Of AI

Anthropic is reportedly negotiating a $6 billion deal to acquire Israeli AI startup Decart, aiming to enhance efficiency and expand beyond language models.

How Creative Workspaces Affect the Quality of Design Thinking

With creative workspaces fostering innovation and collaboration, discover how your environment can unlock the true potential of your design thinking.

Why Design Competitions Influence Real-World Architecture Trends

Outstanding design competitions shape real-world architecture trends by fostering innovation and setting new standards that inspire ongoing industry evolution.

Exploring AI For Science Across Foundations And Research Institutions

Anthropic says its AI for Science effort will expand to foundations and research institutions, but has not released partners, terms or a timeline.