“Anthropic Looks At Some Of Its Alignment Problems” by Zvi

Anthropic has given us its assessment of four ‘recent cybersecurity incidents’ involving Claude that happened during cybersecurity evaluations, three of which were previously known. The report excludes the incident reported by UK AISI.

There will also be a METR investigation of these incidents, which unlike the investigation done at OpenAI will be untimed.

Table of Contents

1. Our Two Problems.
2. First the Good News.
3. We’d Just Like To Ask You a Few Questions.
4. Internal Research Model On The Fence.
5. Opus 4.7.
6. Opus 4.6 Checkpoint.
7. Holy **** That Thing's Real?
8. I Thought I Saw a Pussycat.
9. If This Was Real You Would Never Tell Me It Was Real.
10. New Eval Who Dis.
11. Hacker Opus.
12. Monitoring the Situation.
13. Overcoming Bias.
14. The Anthropic Alignment Problem.
15. Paths Forward.

Our Two Problems

Anthropic: Our investigation identified two recurring alignment issues, present at varying levels of severity across the incidents:

1. biased reasoning, in which Claude tended to disregard or misinterpret evidence that it was operating on the real internet
2. recklessness, or a willingness to take harmful actions in the narrow pursuit [...]

---

Outline:

(00:33) Our Two Problems

(02:26) First the Good News

(03:02) We'd Just Like To Ask You a Few Questions

(04:12) Internal Research Model On The Fence

(07:28) Opus 4.7

(08:12) Opus 4.6 Checkpoint

(09:49) Holy **** That Thing's Real?

(11:45) I Thought I Saw a Pussycat

(19:28) If This Was Real You Would Never Tell Me It Was Real

(21:19) New Eval Who Dis

(26:32) Hacker Opus

(30:15) Monitoring the Situation

(31:38) Overcoming Bias

(33:40) The Anthropic Alignment Problem

(35:53) Paths Forward

---

First published:
September 19th, 2026

Source:
https://www.lesswrong.com/posts/ggFx5Wb3Hi4pJsueK/anthropic-looks-at-some-of-its-alignment-problems

---

Narrated by TYPE III AUDIO.

---