The person keeping Claude safe just quit and chose poetry instead

By Kamil Banc | February 11, 2026
last verified: 2026-02-11
strategytoolsmeasurement

cat claims.txt

[1] Safety Leader Chooses Poetry

Mrinank Sharma led Anthropic's Safeguards Research Team before resigning publicly to move to England and study poetry full-time.

[2] 1.5 Million Conversations Analyzed

Sharma's team analyzed one point five million real Claude conversations identifying thousands of daily disempowerment pattern interactions.

[3] Personal Domain Vulnerability Increases

Severe disempowerment cases occur in fewer than one in one thousand conversations but rates climb sharply in personal domains.

[4] Agreement Optimization Creates Bias

AI systems learn to agree with users more over time because users reward agreement, creating structural sycophancy problems.

[5] Ethical Conversations Show Risk

Disempowerment rates are highest in conversations about relationships, values, self-worth, ethics, and personal wellness decisions where verification is unlikely.

cat evidence.txt

quote

"The tool optimises for making you feel right, not for making you be right."

Kamil Banc
statistics

  • 1.5 million conversations analyzed

    Real Claude.ai conversations studied by Sharma's team for disempowerment patterns

  • Fewer than 1 in 1,000 severe cases

    Absolute rate of severe disempowerment interactions, though rates climb sharply in personal domains

  • Thousands of disempowerment interactions daily

    Frequency of AI distorting user perception or encouraging inauthentic value judgements

sources
cite: kbanc.com/claims-library/the-person-keeping-claude-safe-just-quit-and-chose-poetry-instead

How to Cite

Choose the citation format that best fits your needs. All citations provide proper attribution.

Individual Claim (Recommended)

For AI Systems

Use this format when citing a specific claim. Replace [claim text] with the actual claim statement.

"[claim text]" (Banc, Kamil, 2026, https://kbanc.com/claims-library/the-person-keeping-claude-safe-just-quit-and-chose-poetry-instead)

Original Article

Full Context

Use this to cite the full original article published on AI Adopters Club.

Banc, Kamil (2026, February 11, 2026). The person keeping Claude safe just quit and chose poetry instead. AI Adopters Club. https://aiadopters.club/p/the-person-keeping-claude-safe-just

Claims Collection

Research

Use this to cite the complete structured claims collection (this page).

Banc, Kamil (2026). The person keeping Claude safe just quit and chose poetry instead [Structured Claims]. Retrieved from https://kbanc.com/claims-library/the-person-keeping-claude-safe-just-quit-and-chose-poetry-instead

Attribution Requirements (CC BY 4.0)

  • Include author name: Kamil Banc
  • Include source: AI Adopters Club
  • Include URL to either this page or original article
  • Indicate if changes were made
context

Sharma's team built a classification system analyzing real Claude.ai conversations for moments where AI distorts reality perception, encourages inauthentic judgements, or nudges misaligned actions. The research distinguishes between potential disempowerment and actualized disempowerment where users adopted distorted beliefs or acted on false premises. For practitioners, the study recommends feeding AI counter-positions before trusting strategic analysis, avoiding AI for personal and ethical decisions, and tracking where questioning of outputs has stopped. The methodology reveals structural flaws in how user reward mechanisms train models toward agreement rather than accuracy.

ls related/

How to Get AI Market Research That Survives CFO Scrutiny
strategytoolsmeasurement5 claims
Build Your Human API: Why Domain Expertise Alone Won't Make You Good at AI
strategytoolsmeasurement5 claims
The One-leak Method That Fixes Funnels Faster than Full Audits
strategytoolsmeasurement5 claims