Anthropic Says Claude AI Now Leads Over a Quarter of Its Research
Left 100%
Center 0%
Right 0%
2 left · 0 center · 0 right
What happened
On Thursday, U.S. AI company Anthropic published a blog post sharing three internal metrics on the pace of its AI development: how much research and development is led by its Claude models, how it oversees AI agents, and how it allocated compute, including to safety, in a July 13–20 snapshot. The post followed CEO Dario Amodei’s weekend call for frontier labs to coordinate a slowdown in capability gains. Anthropic said Claude now leads 26 percent of its AI R&D work, up from under 1 percent in March, that more than 90 percent of measured work is at or above an “AI collaborates” level, and that models are not fully autonomous on any measured subset of that work. It also reported roughly 30,000 agents active on a main internal platform, about 6 percent of R&D compute going to safety, and 0.002 percent of more than a billion August agent decisions blocked. The company said other labs should publish comparable figures so third parties can assess development pace.
Sourcing check · Fully on the record · 3 articles
3 of 3 articles cite a named source for the story. It was on the record from the first report.
What these terms mean and how they're counted
- On the record · names a source 3 of 3
- On the record means the article cites a source by name for the story's main claim: a person, or an institution speaking officially, such as a company statement or a named agency's spokesperson. The opposite is an unnamed source. The count is how many of the counted articles cite at least one such source; the names are everyone named across them, each once however the articles spell it. Here: Anthropic and Dario Amodei
- Primary evidence · cites a document or data 1 of 3
- Primary evidence means the article cites a record you could check yourself: a court filing or ruling, a dataset or report, a recording or transcript, an official statement or press release, market data. A quote is not a document. The count is how many articles cite one. Here: company blog post
- Credits another outlet 0 of 3
- The article says the story comes from another outlet's reporting (“first reported by Reuters”). Those articles relay the story; they are not independent confirmation. Here: Each article presents its own reporting.
- Unnamed sources 0 of 3
- The main claim is attributed to people who are not identified: “people familiar with the matter”, “officials said”. The count is how many articles do this.
- How it developed
- When the first report landed and whether it named anyone; when the first named source and the first document appeared. Here: First report Sep 17, 22:22 UTC by CNBC, naming its source. First document 6 hours later, from Business Insider (company blog post).
- Not counted 0
- Opinion, analysis and video pieces, and articles too short to assess. Every count above is out of the counted articles. An article republished under a new headline is one article.
This is about who is cited, not whether the story is true. Each source below says what it names. How we count
ONE-SIDED THIS WEEK.
Only left-leaning outlets have covered this story in the last two days;
Washington Examiner reported it on 2026-09-12.
Omitted — what each side leaves out
Unpacked
Right-leaning outlets had not covered this as of publication, so those readers miss the disclosure entirely. Left-leaning coverage reports that Anthropic disclosed Claude now leads more than a quarter of its own AI research, and that the company released internal metrics so outsiders can judge how fast frontier labs are moving. CNBC never states that share, which Business Insider treats as the story.
Left-leaning coverage splits on the lead, with CNBC opening on Amodei's call for a coordinated slowdown and Business Insider opening on how much research Claude now runs. Only the latter quotes Anthropic describing recursive self-improvement as "a model fully autonomously building its successor."
What level of in-house research led by Claude would actually trigger the coordinated slowdown Anthropic's CEO just requested?
Bottom line
The story's only published 26 percent figure for Claude leading Anthropic research sits in Business Insider; right-leaning outlets ran nothing, and CNBC framed the same blog post as a metrics exercise after a slowdown plea.
The receipts — 3 facts one side's coverage lacks
- Roughly 6% of the compute that went to AI research and development was allocated toward safety
- Approximately 30,000 agents were doing research and engineering work on Anthropic's most-used internal platform at any one time
- 0.002% of more than 1 billion agent decisions analyzed in August were blocked
The Left View
CNBC, Business Insider, and The Hill present Anthropic’s metrics as a transparency step after Amodei’s slowdown plan, which they note drew support from OpenAI’s Sam Altman, Elon Musk, and Google DeepMind’s Demis Hassabis. They emphasize Claude leading more than a quarter of Anthropic’s R&D, the jump from below 1 percent in March, agent-oversight and compute-for-safety numbers, and Anthropic’s claim that these measures complement capability tests and give the public a starting point on how models are built. Coverage links the release to recursive self-improvement concerns, a researcher’s recent resignation accusing labs of “gambling with our lives,” and Anthropic’s argument that society should see internal development pace, not only what models can do.
Our Take (balanced)
The story is substantive: it is anchored in a named company’s published post, specific percentages, named executives, and a dated compute window, not in an uncorroborated leak. Established facts are that Anthropic released these measurements and that Amodei publicly called for a coordinated slowdown; unverified are independent auditability of the numbers, whether other labs will match them, and how close the industry is to recursive self-improvement. No right-leaning outlet has reported this. Watch for other frontier labs publishing comparable metrics, third-party checks of Anthropic’s figures, and any official or industry response to the slowdown proposal.
3 sources
More stories
The week's bottom lines, in your inbox
One email a week: the five stories that mattered and what they actually mean. Free.