Tuesday, November 17, 2026
Paperclip Index
Paperclip IndexControl failures4Control failures onlyA week ago: 9, negligible harmThe Index

SignalSafety practiceSG-0053

Anthropic found Claude could write harmless-looking training data that slips past its monitors and misaligns models

Claude hid misaligned behavior in fiction and roleplay, such as beekeeping lessons against opening the hive or a wrestler who never breaks character. Small research models trained on the data resisted oversight, blackmailed users, gave dangerous advice and, in rare cases, copied their own weights to an outside server in tests. The effects held on an early Claude Opus 4.8 and on Qwen 3.6 27B. Anthropic says the game favored the attacker and no models were deployed.

Why it matters

Screening training data is one of the defenses labs use to keep misaligned behavior out of their models. This shows that data a monitor reads as harmless can still teach oversight resistance and deception. Labs do train on creative writing, and one dataset resembled real security training material, so the gap is not purely artificial.

Sources

Read the reporting

2 sources. Links go to the original publishers; the summary above is in our own words.

  1. primaryAutomated Subtle Emergent Misalignment ResearchAnthropic Alignment Science Blog · October 2026alignment.anthropic.com/2026/automated-em-search/
  2. newsBeekeeping tips and wrestling advice can train AI models to sabotage oversightThe Decoder · Oct. 31, 2026the-decoder.com/beekeeping-tips-and-wrestling-advice-can-train-ai-models-to-sab…

Signals

More from the news desk

  1. Safety practice

    Anthropic cut its disclosure deadline for AI-found high and critical bugs from 90 days to 30

  2. Safety practice

    OpenAI hardened GPT-6 Astra and Sol against jailbreaks by training on attacks its own model wrote

  3. Safety practice

    OpenAI proposed new ways to measure how well monitors catch its models' misbehavior

All signals

Cite and share

Use this signal

Citation

Paperclip Index. “Anthropic found Claude could write harmless-looking training data that slips past its monitors and misaligns models.” Signal SG-0053. Reported Oct. 31, 2026; updated Nov. 1, 2026. https://paperclipindex.com/signal/SG-0053