Anthropic called on the world’s leading artificial-intelligence labs Thursday, June 4, to consider a coordinated slowdown or temporary pause in building the most advanced AI systems, warning that the technology is approaching the point where it could improve itself without human help. The recommendation came in a report from the company’s research arm, the Anthropic Institute, written by Marina Favaro, who leads its internal research, and co-founder Jack Clark, the company’s head of policy.

The company said the world having the option to slow or temporarily pause frontier AI development would “likely be a good thing,” arguing it would give governments, institutions, and safety researchers time to catch up with how fast the technology is advancing. It is a striking message from a company that is itself one of the fastest-moving developers in the field.

At the center of the warning is a concept researchers call recursive self-improvement. In plain terms, it describes the moment an AI system becomes capable of improving itself, or designing and building its own successor, without much human involvement. The report said models are showing early signs of moving in that direction, a threshold the company said could bring major disruption if it is crossed before society is ready.

Anthropic backed the warning with data about its own operations. The company said more than 80% of the code merged into its systems is now written by its Claude models, and that its engineers ship roughly eight times as much code per quarter as they did in the years from 2021 through 2025. In other words, the company says its AI is already accelerating the pace at which AI itself is built.

The report was careful to add limits. Anthropic said the industry is not yet at recursive self-improvement, and that such a future is not inevitable. But it warned the moment could arrive sooner than most institutions are prepared for. In comments to BBC News, Clark said AI reaching the point of writing its own code fully could be possible within about two years.

The proposal faces an obvious problem, which Anthropic acknowledged: if a single company slowed down on its own, competitors would simply race ahead. For that reason, the company argued any pause would have to be coordinated globally and verifiable. The Anthropic Institute said it plans to research and develop systems that would let frontier developers confirm rivals have actually stopped, and ensure no bad actor uses a coordinated slowdown to quietly pull ahead.

The call lands at a moment of intense commercial pressure across the industry. Anthropic recently completed a funding round that valued the company at nearly $1 trillion and has filed confidential paperwork to begin the process of going public. Its run rate, a measure startups use to project annual revenue from recent sales, is on track to reach about $50 billion in annualized revenue by the end of this month, up from roughly $9 billion at the end of 2025. The company has emerged as a front-runner against OpenAI, the maker of ChatGPT, which is also expected to pursue a public listing.

That commercial position fuels a long-running criticism. Anthropic has emphasized AI safety since its founding, but skeptics, including venture capitalist David Sacks, have argued that its policy advocacy is designed in part to slow the progress of competitors. A public call for rivals to consider pausing is likely to renew that debate, even as Anthropic frames the recommendation as a matter of public risk rather than competitive advantage.

The stakes extend across an industry that is spending hundreds of billions of dollars on data centers, chips, and talent. A coordinated pause would affect the entire race, from the largest technology companies to the startups built on their models. Anthropic’s central recommendation is not that development stop now, but that governments and labs preserve the ability to pause and build the infrastructure that would make such a pause credible if it became necessary.

The report arrives the same week that a bipartisan group of House lawmakers unveiled draft legislation to regulate AI, underscoring how questions about the technology’s speed and safety are moving to the center of policy discussions. Anthropic’s proposal adds a prominent industry voice to that conversation, with the company arguing that the option to slow down, backed by ways to verify it, should exist before the technology reaches a point where stopping becomes far harder.

JBizNews Desk — Artificial Intelligence

© JBizNews.com All Rights Reserved. Reproduction or distribution without written permission is prohibited.