Anthropic's Claude Opus 4.7 can identify writers from unpublished text across multiple genres, according to experiments conducted by Vox journalist Kelsey Piper in April 2026. Piper pasted 125 words of an unpublished political column into the model and received her own name back—despite testing in Incognito mode, through the API, and on a friend's computer. The discovery suggests that anyone who has written prolifically under their real name may have lost meaningful anonymity.
Claude Opus 4.7 Identified Writers Across Diverse Unpublished Content
Piper's experiments revealed that Claude Opus 4.7 could identify her writing across radically different formats and genres:
- A school progress report about a student's Pokémon essays (entirely outside her published work)
- A movie review of a 1942 WWII comedy she had never publicly reviewed
- 500 words of unpublished fiction
- A 15-year-old college application essay
In contrast, ChatGPT and Gemini mostly guessed incorrectly on the same texts, suggesting Opus 4.7 has unique capabilities in stylometric analysis—the study of linguistic patterns that reveal authorship.
Privacy Implications for Whistleblowers and Pseudonymous Authors
The model's ability to identify writers from unpublished text has significant implications for privacy and anonymity:
- Whistleblowers attempting to publish anonymously could be identified
- Pseudonymous authors may be connected to their real identities
- Journalists protecting sources face new challenges
- Anyone separating professional and personal writing may be at risk of de-anonymization
Piper noted in The Argument that the discovery means "anyone who has written prolifically under their real name has probably lost meaningful anonymity."
Context: Opus 4.7's April 2026 Release
Anthropic released Claude Opus 4.7 in April 2026, emphasizing improvements in coding, visual tasks, and cybersecurity guardrails. CNBC described it as "an AI model that is less risky than Mythos." However, the privacy implications of its advanced pattern recognition capabilities were not part of the initial announcement.
The Hacker News discussion of Piper's findings received 128 points and 64 comments, with developers examining the technical mechanisms behind stylometric fingerprinting and discussing potential safeguards.
Key Takeaways
- Claude Opus 4.7 identified Kelsey Piper from 125 words of unpublished text, confirmed through multiple independent tests in Incognito mode and via API
- The model successfully identified her writing across diverse genres including fiction, school reports, movie reviews, and a 15-year-old college essay
- ChatGPT and Gemini failed where Opus 4.7 succeeded, indicating unique stylometric analysis capabilities
- The discovery threatens anonymity for whistleblowers, pseudonymous authors, and anyone attempting to separate personal and professional writing
- Anthropic released Claude Opus 4.7 in April 2026 with a focus on coding and cybersecurity, but privacy implications were not prominently disclosed