Analyzing the News: Anthropic's IPO filing

Thirteen outlets covered Anthropic's filing to go public. Fíor read three of them as AI, and flagged five for inflated significance.

Fíor is a natural language classifier that predicts whether a piece of text was more likely written by a person or by AI.

On 26 October, Anthropic, the company behind the Claude chatbot, published its paperwork for a stock market listing. It plans to list on Nasdaq under the ticker ANT, but has not yet set a price or the number of shares. The filing showed revenue of $16.3 billion in the first half of 2026, up from about $1.16 billion a year earlier. It also showed a net loss of $80.7 billion, though $72.4 billion of that was a non-cash accounting charge caused by the company's own rising valuation. Its operating loss was $693 million.

The filing also set out commitments worth hundreds of billions of dollars for computing power over the coming years, and gave its founders control of shareholder votes after the listing. Reports have suggested the offering could value the company at $1.5 trillion to $2 trillion. ABC News counted 90 pages of the filing given over to risks, including Anthropic's own warnings about what advanced AI could do.

The outlets were Reuters, AP, Bloomberg, the Guardian, TechCrunch, Fortune, Forbes, CBS News, Business Insider, ABC News, Axios, The Verge and WIRED. That is two wire services, the business press, American broadcasters, the Guardian and technology outlets. Only part of Bloomberg's article was scanned. WIRED sits behind a paywall, and its scan also picked up subscription and author-page text along with the opening of the article.

The result

Ten read as human: Reuters, AP, Bloomberg, TechCrunch, Fortune, Forbes, Business Insider, Axios, The Verge and WIRED.

Three read as AI: the Guardian, CBS News and ABC News.

Fortune was a close call. It came down only just on the human side, and Brath marked it not confident. Bloomberg and Axios also read as human but not confidently, and ABC News read as AI but not confidently. CBS News was the clearest AI result, and the only one where Brath also said the article reads like AI.

This is the ninth story scanned with Brath, Fíor's new model. CBS News has now read as AI in four stories running where it ran its own reporting. The Guardian has read as AI in seven of the eleven stories where it ran its own reporting.

The flags came back

For the last few stories, Brath's writing-habit flags had gone quiet. In the Barclays and Lula pieces, almost nothing was flagged. This time five articles were flagged for inflated significance, Brath's marker for writing that leans hard on how much something matters: the Guardian, CBS News, Bloomberg, Forbes and The Verge. Forbes was also flagged for dense, abstract phrasing.

The flags did not line up with the verdicts. Forbes and The Verge drew flags and still read clearly as human. ABC News read as AI and drew none. The flags and the verdict are separate signals, and here they told different stories.

The obvious explanation is the subject. Anthropic's filing describes its own work in very large terms, and coverage that quotes it picks up the language. The Verge's headline quoted the company's plan to "revolutionize every area of human endeavor". That does not explain everything, though. Fortune's headline quoted the filing's "age of miracles", and Fortune drew no flag at all.

Why IPO coverage reads like this

Every outlet worked from the same long document, published on the same day. Most pulled out the same numbers: the revenue, the headline loss and the accounting charge behind it. Where they differed was in what they treated as the story. Reuters and the Guardian made it a test of the whole AI investment boom. AP, ABC News and CBS News led on the company warning about the risks of its own technology. TechCrunch focused on who will control the company after it lists.

Some of these pieces may have had AI help, with a first draft, a rewrite or a final polish. Or IPO coverage has become formulaic enough that a reporter and a model would write it the same way. Fíor cannot tell those apart on its own.

Anthropic's filing promises an age of miracles and spends 90 pages on what could go wrong. Five of the articles about it drew Fíor's flag for inflated significance, the most in any story so far. It is hard to write about a company that describes its own work in those terms without some of the language rubbing off.

What Fíor is

Fíor is a machine learning classifier, currently in beta and available as a browser extension. It gives a prediction, not a verdict. It cannot see who sat at the keyboard, and nothing here is a claim about any journalist or publication. Reading as AI and being written by AI are not the same thing.

Treat it as a signal, not proof, and check the pages you read yourself.

Try the browser extension at Fíor.

Analyzing the News is a series where we run one major story through Fíor and see what comes back.