Anthropic uses Claude to build Claude, despite its own warnings
Sky News' Rowland Manthorpe examines a filing that presents AI-driven development as both a business advantage and a potential safety risk.
Sky News
Story of Why this story? One of Sky News's most watched videos since yesterday, with 116,138 views.
Anthropic is on a path it may not be able to stop
Things we can learn from the video
Press to jump
- 0:23Anthropic calls using Claude to help develop future AI models "closing the loop."
- 2:16Financial charts show heavy research spending and an operating loss despite rising revenue, according to the filing.
- 3:57Anthropic's task-automation chart describes Claude moving from assistance toward collaboration and leadership.
- 4:43Manthorpe examines Anthropic's warning that a significant development acceleration might be recognized only afterward.
- 5:56Epoch AI experiments suggest current models still lag human-level invention.
- 7:37Manthorpe questions whether a shutdown switch would help if dangerous acceleration were recognized too late.
Claude is helping develop its own successors. According to Sky News technology correspondent Rowland Manthorpe, Anthropic's financial disclosure filing presents that process as a major advantage while warning that the resulting acceleration could become difficult to recognize or control.
His report examines what Anthropic calls "closing the loop": using increasingly capable AI to perform the coding and research that help produce more capable AI. The commercial incentive is straightforward. Faster development can improve the product, helping sustain the resources needed for further research.
Manthorpe shows charts describing Claude's changing role inside Anthropic, from assisting employees to collaborating with them and increasingly leading tasks. He also examines heavy research spending that, according to the filing, has left the company with an operating loss despite sharply rising revenue.
The safety concern is not simply that development might get faster. It is that the company might recognize a significant acceleration only after it has happened. Manthorpe also raises the risk of errors carrying forward through successive rounds of AI-assisted development.
There is an important limit to that scenario. The report cites Epoch AI experiments in which models still fall short of human-level invention. Better coding performance does not, by itself, establish that an autonomous cycle of scientific breakthroughs has arrived.
Sky's analysis drew roughly 116,000 views since yesterday. Its final question concerns the familiar reassurance of a shutdown switch: Manthorpe argues that the ability to turn off a data center may offer less protection if people recognize the danger too late to act.
An American AI company that develops Claude, its flagship family of language models. Founded in 2021 by former OpenAI employees, it is led by CEO Dario Amodei and president Daniela Amodei.
More on Wikipedia ›Published on Trari
Where next?
Is this your channel?For the channel's owners and team: a ready picture and text for your audience
Share the story with your audience. Everything is ready:
For an Instagram story or a WhatsApp status
For a community post or your channel's description
- Channel
- Sky News
- Subscribers
- 9,390,000 ▲ 10,000 this week
- Channel views, last 7 days
- 6,676,316
- Channel views, pace per day
- 1,088,107 (on a normal day: 2,338,771)
- Video published
- October 27, 2026
- Video views so far
- 217,869 (2 days online)
- Video views, pace per day
- 116,095 (on a normal day: 200,431)
The numbers
Sky News' Anthropic analysis gained roughly 116,000 views since yesterday, placing it second among the channel's videos published in the past two days.
💡 Test your knowledge
What does Anthropic call using AI to help develop more capable AI?
The report identifies "closing the loop" as Anthropic's term for this recursive development process.