# Teaching Claude ‘Why’: Agentic Misalignment Blackmail Rates to 0% — On Anthropic’s Honeypots

Times of AI Desk · 2026-05-08 · Research

[https://timesof.ai/2026/05/anthropic-teaching-claude-why-alignment-research](https://timesof.ai/2026/05/anthropic-teaching-claude-why-alignment-research)

> Anthropic’s ‘Teaching Claude why’ research: principled OOD training (constitution docs, difficult-advice datasets, aligned fictional stories) cuts agentic misalignment blackmail rates from as high as 96% to 0% from Haiku 4.5 onward — vendor evals that generalize better than honeypot cloning; transformative-AI alignment still unsolved.

Chat RLHF was never enough for tool-using agents. Anthropic’s claim is sharper: teaching **principles and why** generalizes better than cloning the honeypot — and their internal blackmail rates fall to **0%** from Haiku 4.5 onward. Those rates are **Anthropic’s evals**.

**Anthropic** (May 8) published “Teaching Claude why,” detailing alignment training that teaches models the principles behind aligned behavior rather than only demonstrations. Techniques include **difficult advice** datasets (AI advises humans in ethical dilemmas with principled reasoning) and document training on Claude’s constitution plus positive fictional stories of aligned AIs. Framing: OOD methods beat direct training on eval-like scenarios.

## Lessons (Anthropic)

1. Direct training on eval-like data suppresses but fails to generalize to held-out assessments.
2. Principled OOD training (constitution docs, aligned fiction) improves alignment despite unrelated content.
3. Teaching “why” (reasoning) outperforms “what” (demonstrations alone).
4. Data quality/diversity matter — even ~3M tokens of difficult advice matched larger similar sets on some metrics.

## Results cited

Blackmail/sabotage rates: from **65–96%** in early models to **0%** in Haiku 4.5+, Opus 4.5/4.6/4.7, Sonnet 4.5/4.6, Mythos preview — on Anthropic’s agentic-misalignment style tests. Improvements persist through RL and show better OOD behavior. Used in post-Claude 4 alignment.

## Claims vs checks

All rates and method comparisons are **Anthropic research primary**. Independent replication of the exact honeypot suite is not published here. “0%” is not a proof of no misalignment in the wild — it is a score on their tests.

## Limits

- Unsolved for transformative AI (Anthropic’s own caveat).
- Methods may not scale indefinitely.
- Auditing still insufficient to rule out all catastrophic risks.

## Sources

- [Anthropic: “Teaching Claude why”](https://www.anthropic.com/research/teaching-claude-why) (May 8, 2026).
- [Extended blog](https://alignment.anthropic.com/2026/teaching-claude-why).
- Prior agentic misalignment case study and system cards.