corpus.blog
56,966 blogs · 4,417,066 posts Claim your blog
Record · corpus.blog/posts/01a07d5d-9926-72c1-bdf5-2e8dc61caca4

SOTA alignment assessments don’t strongly update us against misalignment

blog.redwoodresearch.org · published 31 July 2026

A record is what was published and who pointed at it. The text of the post is not held here: read it at the source, or at its Wayback capture.

Read the post at blog.redwoodresearch.org Wayback capture
1blogs citing
First cited
25 Sept 2026
Most recent
25 Sept 2026
Rank this month
not ranked
Links from this post
60
Words captured
9,672

Cited by

Blog
In the post
Date
LessWrong
The Geopolitics of an AI Treaty Read at the source ↗
25 Sept 2026

Links from this post

Link
Host
https://www-cdn.anthropic.com/3edfc1a7f947aa81841cf88305cb513f184c36ae/Alignment%20Risk%20Update_%20Claude%20Mythos%20Preview%20(Redacted,%20April%2010
www-cdn.anthropic.com
https://www.lesswrong.com/posts/HE3Styo9vpk7m8zi4/evhub-s-shortform?commentId=Sy77So6NXQ9nvbG7M
lesswrong.com
https://www.lesswrong.com/posts/nAsMfmxDv6Qp7cfHh/fabien-s-shortform?commentId=c9vzjifqSFW585rKY
lesswrong.com
https://metr.org/blog/2026-05-19-frontier-risk-report/
metr.org
https://substackcdn.com/image/fetch/$s_!oynz!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F8ba512c3-3828-48ee-89cd-22354eaa0ebc_1772x989.png
substackcdn.com
https://substackcdn.com/image/fetch/$s_!CaGP!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F7e28d343-2d97-4be3-8d4e-5d7bf08a6fca_1362x411.png
substackcdn.com
https://substackcdn.com/image/upload/f_auto,q_auto/v1/mirroredImages/4ae97320c14eea3e3e7d91954b1e14674b55d25fb4b9167e795ad4616b564ed2/zgwihixkrxk11wbktmm1
substackcdn.com
https://substackcdn.com/image/upload/f_auto,q_auto/v1/mirroredImages/78b1c44b69458f7318f0b0f3c57421ff77fc80671cfeff000f6559005cbb3b73/jhbm2oglyzr6tqlmwujs
substackcdn.com
https://substackcdn.com/image/fetch/$s_!7pTf!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe0ed8b97-a58e-4a41-a852-9d87529a375c_1366x504.png
substackcdn.com
https://substackcdn.com/image/fetch/$s_!4GXU!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fa1fc44c7-bdf1-477e-9079-59121ebb9c38_1778x544.png
substackcdn.com
https://substackcdn.com/image/upload/f_auto,q_auto/v1/mirroredImages/cfba3ba1032fe0ae01d9ea6054419190763bfd321fab08aed66505a1a00d7186/vaexztzt5p8juqm4bk4a
substackcdn.com
https://alignment.openai.com/metagaming/
alignment.openai.com
https://deploymentsafety.openai.com/gpt-5-5/cot-monitorability
deploymentsafety.openai.com
https://deploymentsafety.openai.com/gpt-5-5/cot-controllability
deploymentsafety.openai.com
https://www-cdn.anthropic.com/79c2d46d997783b9d2fb3241de43218158e5f25c.pdf
www-cdn.anthropic.com
https://www-cdn.anthropic.com/08ab9158070959f88f296514c21b7facce6f52bc.pdf
www-cdn.anthropic.com
https://www.anthropic.com/research/teaching-claude-why
anthropic.com
https://www-cdn.anthropic.com/14e4fb01875d2a69f646fa5e574dea2b1c0ff7b5.pdf
www-cdn.anthropic.com
https://www.lesswrong.com/posts/Jtk6rfjpABaAyyhsH/alexa-pan-s-shortform?commentId=nZk7wJt9xS77AFgrr
lesswrong.com
https://ai.meta.com/static-resource/muse-spark-safety-and-preparedness-report/
ai.meta.com
Sidestepping Evaluation Awareness and Anticipating Misalignment with Production Evaluations Read at the source ↗
alignment.openai.com
https://www-cdn.anthropic.com/3edfc1a7f947aa81841cf88305cb513f184c36ae.pdf
www-cdn.anthropic.com
https://www.lesswrong.com/posts/qgehQxiTXj53X49mM/sonnet-4-5-s-eval-gaming-seriously-undermines-alignment?_ga=2.163787069.303343724.1764174183-1438369982.1737922106
lesswrong.com
https://www-cdn.anthropic.com/d00db56fa754a1b115b6dd7cb2e3c342ee809620.pdf
www-cdn.anthropic.com
https://arxiv.org/abs/2502.02180
arxiv.org
https://deploymentsafety.openai.com/gpt-5-5/research-category-update-sandbagging
deploymentsafety.openai.com
https://cdn.sanity.io/files/4zrzovbb/website/037f06850df7fbe871e206dad004c3db5fd50340.pdf
cdn.sanity.io
https://www.anthropic.com/research/auditing-hidden-objectives
anthropic.com
https://substackcdn.com/image/fetch/$s_!-ayO!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fceab08f7-4ccb-4a02-90d3-176cb7470ccf_968x678.png
substackcdn.com
https://substackcdn.com/image/upload/f_auto,q_auto/v1/mirroredImages/b64fd182957e30cd4d052e21cdacad6575735ed5a1ed3fbef8923824eac1274e/bedwqvgtkm25kwoc94ut
substackcdn.com
Risk reports need to address deployment-time spread of misalignment Read at the source ↗
blog.redwoodresearch.org
https://alignment.anthropic.com/2026/auditbench/
alignment.anthropic.com
https://substackcdn.com/image/fetch/$s_!N1pi!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fbd269296-28b8-449d-b65a-cf140621806a_2048x901.png
substackcdn.com
https://substackcdn.com/image/upload/f_auto,q_auto/v1/mirroredImages/c8d21f3c00b9455c6b36db97be6163f2567798f9344f262d8209a4a2d55bedd1/jkn8xxpwr8fmb3ftfmti
substackcdn.com
https://www.lesswrong.com/posts/LhxHcASQwpNa3mRNk/untrusted-smart-models-and-trusted-dumb-models
lesswrong.com
https://www.lesswrong.com/posts/TTFsKxQThrqgWeXYJ/how-might-we-safely-pass-the-buck-to-ai
lesswrong.com
https://www.lesswrong.com/posts/4QRvFCzhFbedmNfp4/to-be-legible-evidence-of-misalignment-probably-has-to-be
lesswrong.com
https://substackcdn.com/image/fetch/$s_!nOfc!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F056a4762-ab30-401c-8c9e-0f6eb57e5d94_1156x312.png
substackcdn.com
https://substackcdn.com/image/upload/f_auto,q_auto/v1/mirroredImages/ec5ff3a1f94c30ef2769e8bef24c1a749a29e8da54af76c0163c09d399200a62/chbrpwbgermhsjcjddtj
substackcdn.com
https://substackcdn.com/image/upload/f_auto,q_auto/v1/mirroredImages/6387af7ebef506388bfe9f949124e94c08cccb848faaaea25c69933d81198279/mobfwipdzgfhdgjwr3ve
substackcdn.com
https://substackcdn.com/image/fetch/$s_!4Kgd!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F80fb15d7-ee04-40d0-b211-9fe235bcafd4_1156x215.png
substackcdn.com
https://substackcdn.com/image/fetch/$s_!TM9G!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3ab6dcaa-f1b4-445f-b6b6-7f356e6664fe_992x350.png
substackcdn.com
https://substackcdn.com/image/upload/f_auto,q_auto/v1/mirroredImages/oirrSj3itFLSyscW8/79ae51136f24e837d4f5b7c3d62c6e6f8ba359b145f6228d329625666351a1bc/y1m2s9mzocbuma3zmmiv
substackcdn.com
https://substackcdn.com/image/fetch/$s_!OVfL!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fd8b87f4f-4380-47ab-8146-72e94746c88d_918x296.png
substackcdn.com
https://substackcdn.com/image/upload/f_auto,q_auto/v1/mirroredImages/34d9b50a45426f124e061c961bcffd094025c9fb32df777fdd1ddab0ebfbacab/zdnvre2mr4wh996lftdh
substackcdn.com
https://substackcdn.com/image/fetch/$s_!Lisd!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fe95fed53-581d-4358-bdf5-fe2ee13b1739_1078x314.png
substackcdn.com
https://substackcdn.com/image/upload/f_auto,q_auto/v1/mirroredImages/cf03008111943b83bb7831e194984586e478e60f83917948ce50007fff53fbb8/qreze9sj0xvoc3z5jtjj
substackcdn.com
https://www.lesswrong.com/posts/oirrSj3itFLSyscW8/sota-alignment-assessments-don-t-strongly-update-us-against
lesswrong.com
https://substackcdn.com/image/fetch/$s_!c5fE!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F0cbbfbc0-4eb0-4ae7-9e9f-3ba4913d363c_980x506.png
substackcdn.com
https://substackcdn.com/image/upload/f_auto,q_auto/v1/mirroredImages/7d0eb5f75030374114c5818a25098e0e29553144d2f9ab34fc0110959b6fd049/x9glvuuzodafa27ilsaf
substackcdn.com
https://substackcdn.com/image/fetch/$s_!o720!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F9696aea0-934b-4b55-a74b-7ed5a5aaa8f1_968x432.png
substackcdn.com
https://substackcdn.com/image/upload/f_auto,q_auto/v1/mirroredImages/774f86eb41d652aee879f052c4e220f4375bb9f52bdccc8efec119f4ae8ea65e/d9qf4gnfvtbsvuewui7i
substackcdn.com
https://substackcdn.com/image/fetch/$s_!OcRI!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2Fdb895f18-5e21-4c11-8b8a-1a9f173896d0_822x270.png
substackcdn.com
https://substackcdn.com/image/upload/f_auto,q_auto/v1/mirroredImages/df52ec759e013fbfc035815aa89de76f3e026dd468d563e85eab54c8b2487e49/jwjrqlla8l011ufhynns
substackcdn.com
https://substackcdn.com/image/fetch/$s_!tcPd!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F3e090496-889e-4ea4-810b-a59d370a78e5_829x349.png
substackcdn.com
https://substackcdn.com/image/upload/f_auto,q_auto/v1/mirroredImages/940c8bd5df21483b9ffd19559fbe49500bbad10300ed45eca56a1e8be13eb934/f8tpzorkl7wod3exegcz
substackcdn.com
https://substackcdn.com/image/fetch/$s_!0LBL!,f_auto,q_auto:good,fl_progressive:steep/https%3A%2F%2Fsubstack-post-media.s3.amazonaws.com%2Fpublic%2Fimages%2F4203a2d1-267f-4539-a2e7-969b22ddf86f_1663x901.png
substackcdn.com
https://blog.redwoodresearch.org/p/anthropic-repeatedly-accidentally
blog.redwoodresearch.org
https://www.lesswrong.com/posts/aTcsN5ZZDnMFJvRiG/models-may-behave-worse-when-eval-aware
lesswrong.com
https://www.google.com/search?q=While+many+of+our+tests+are+recognizably+unrealistic%2C+there+is+significant+variation.+We+believe+that%2C+on+the+basis+of+all+of+the+sources+of+evidence+discussed+above%E2%80%94including+our+heavy+use+of+real+data+from+pilot+internal+and+external+use%E2%80%94this+risk+is+adequately+managed+at+the+current+level+of+capability&rlz=1C5OZZY_enUS1214US1216&oq=While+many+of+our+tests+are+recognizably+unrealistic%2C+there+is+significant+variation.+We+believe+that%2C+on+the+basis+of+all+of+the+sources+of+evidence+discussed+above%E2%80%94including+our+heavy+use+of+real+data+from+pilot+internal+and+external+use%E2%80%94this+risk+is+adequately+managed+at+the+current+level+of+capability&gs_lcrp=EgZjaHJvbWUyBggAEEUYOdIBBzM1NmowajeoAgCwAgA&sourceid=chrome&ie=UTF-8
google.com