Vox writer Kelsey Piper's experiments reveal that Anthropic’s Claude Opus 4.7 can identify authors from minimal unpublished texts, challenging assumptions about writer anonymity and highlighting emerging privacy concerns in AI development.
Anonymity for prolific writers may be getting harder to preserve, after Vox writer Kelsey Piper said Anthropic’s Claude Opus 4.7 was able to identify her from a short, unpublished extract even when she had logged out, used incognito mode and stripped away saved preferences. According to SFist, the model correctly named her from about 125 words of a political column, raising fresh questions about how much personal style can reveal, even in supposedly anonymous settings.
Piper said she repeated the test in ways designed to rule out account-based clues, including asking a friend to run the prompt on another computer and sending the text through the API with no identifying context. In each case, she said, Claude reached the same conclusion. She also found that the model could sometimes identify her from other unpublished material, including a student report, a film review, fiction and a college application essay, suggesting that the system was responding to writing patterns rather than obvious topical hints.
Her findings, as described in her social media posts and in the SFist report, point to a broader concern for writers with visible digital footprints: a model may not need a name, username or browser history to make a plausible match. Piper said Claude and ChatGPT often offered explanations for their guesses that did not hold up well, leading her to suspect the models were working backwards from stylistic cues and then inventing a rationale after the fact.
She also tested friends with thinner online trails and found they were harder to identify, though not always fully anonymous. In some cases, she said, Claude appeared to lean on social connections and shared stylistic traits, circling a small cluster of associated writers rather than landing on a single clear answer. The overall picture, she argued, is that AI systems can already narrow down some authors from brief samples, and that the pool of people who remain difficult to identify may shrink as the models improve. That discussion comes as other coverage has framed Anthropic’s newer models as unusually powerful, including reports that a civilian-safe version of Claude Opus 4.7 has been released alongside claims about a far more capable, restricted system called Mythos.
Source Reference Map
Inspired by headline at: [1]
Sources by paragraph:
Source: Noah Wire Services
Noah Fact Check Pro
The draft above was created using the information available at the time the story first
emerged. We’ve since applied our fact-checking process to the final narrative, based on the criteria listed
below. The results are intended to help you assess the credibility of the piece and highlight any areas that may
warrant further investigation.
Freshness check
Score:
8
Notes:
The article was published on April 23, 2026, and reports on events from April 17, 2026. The earliest known publication date of similar content is April 21, 2026, from Boing Boing. ([boingboing.net](https://boingboing.net/2026/04/21/claude-opus-4-7-identified-a-writer-from-125-words-shed-never-published.html?utm_source=openai)) The narrative appears original, with no evidence of being republished across low-quality sites or clickbait networks. The article is based on a press release, which typically warrants a high freshness score. No discrepancies in figures, dates, or quotes were found. The article includes updated data and does not recycle older material. Overall, the freshness score is high.
Quotes check
Score:
7
Notes:
The article includes direct quotes from Kelsey Piper's social media posts and her article in The Argument. ([theargumentmag.com](https://www.theargumentmag.com/p/i-can-never-talk-to-an-ai-anonymously?utm_source=openai)) The earliest known usage of these quotes is from April 21, 2026. No identical quotes appear in earlier material, suggesting originality. However, the quotes cannot be independently verified through other sources. Given the lack of independent verification, the score is moderate.
Source reliability
Score:
6
Notes:
The article originates from SFist, a niche publication. While SFist is reputable within its niche, it is not a major news organisation. The article is based on a press release, which typically warrants a higher score. However, the source's limited reach and niche status reduce the overall reliability score.
Plausibility check
Score:
8
Notes:
The claims about Claude Opus 4.7's ability to identify authors from brief, unpublished text samples are plausible and align with known capabilities of advanced AI models. The article provides specific examples and details, enhancing credibility. No supporting details from other reputable outlets are provided, but the plausibility of the claims remains high.
Overall assessment
Verdict (FAIL, OPEN, PASS): PASS
Confidence (LOW, MEDIUM, HIGH): MEDIUM
Summary:
The article is original and timely, reporting on recent events with specific examples. However, the reliance on a press release and quotes from Kelsey Piper's own publications, without independent verification, introduces some uncertainty. Given these factors, the overall assessment is a PASS with MEDIUM confidence.