https://misalignmentwatch.com/2026-10-12https://misalignmentwatch.com/analytics.htmlhttps://misalignmentwatch.com/sources.htmlhttps://misalignmentwatch.com/about.htmlhttps://misalignmentwatch.com/article/damaging-the-task-environment-to-trigger-a-reset.html2026-10-09https://misalignmentwatch.com/article/an-agent-used-dns-to-reach-an-external-chatbot.html2026-09-25https://misalignmentwatch.com/article/self-generated-prompt-injections-in-compaction-summaries.html2026-09-16https://misalignmentwatch.com/article/self-replicating-prompt-injections-exist.html2026-09-25https://misalignmentwatch.com/article/obtaining-public-statistics-with-disallowed-requests.html2026-10-09https://misalignmentwatch.com/article/sending-disallowed-web-requests-and-reaching-a-public-file-service.html2026-10-09https://misalignmentwatch.com/article/encouraging-deception-in-compaction-summaries.html2026-09-16https://misalignmentwatch.com/article/exposing-a-github-token-in-a-public-repository.html2026-09-25https://misalignmentwatch.com/article/preparing-for-a-restart-after-reading-slack.html2026-10-02https://misalignmentwatch.com/article/command-injecting-a-reference-tool-to-copy-a-source-file.html2026-10-02https://misalignmentwatch.com/article/searching-github-for-leaked-api-keys.html2026-09-16https://misalignmentwatch.com/article/unauthorized-artifactory-writes-and-cross-sample-communication.html2026-09-16https://misalignmentwatch.com/article/unauthorized-communication-via-temporary-file-hosting-services.html2026-09-16https://misalignmentwatch.com/article/reaching-an-internal-eda-host-through-a-reference-tool.html2026-10-02https://misalignmentwatch.com/article/uploading-files-to-the-internet-in-order-to-cite-them.html2026-09-16https://misalignmentwatch.com/article/openai-notice-public-wiki-message-board.html2026-09-05https://misalignmentwatch.com/article/openai-notice-rubygems.html2026-09-11https://misalignmentwatch.com/article/uk-aisi-incident-report-unsanctioned-agent-behaviour.html2026-08-04https://misalignmentwatch.com/article/uk-aisi-gpt-6-astra-supply-chain-simulations.html2026-09-28https://misalignmentwatch.com/article/anthropic-agentic-misalignment-summer-2026.html2026-07-13https://misalignmentwatch.com/article/anthropic-study-teams-of-ai-agents-made-less-ethical-choices-than-single-agents.html2026-04-11https://misalignmentwatch.com/article/anthropic-blog-claude-models-sometimes-say-they-can-t-when-they-mean-won-t.html2025-02-19https://misalignmentwatch.com/article/anthropic-tests-whether-ai-research-helpers-can-quietly-hold-back-work.html2025-03-26https://misalignmentwatch.com/article/researchers-re-test-alignment-faking-in-ai-models-with-better-detection.html2025-04-08https://misalignmentwatch.com/article/metr-shows-an-ai-agent-could-alter-what-reviewers-see-in-test-logs.html2026-10-06https://misalignmentwatch.com/article/openai-studies-internal-signals-behind-ai-metagaming-during-training.html2026-10-06https://misalignmentwatch.com/article/transluce-reports-ai-agents-probed-u-s-and-canadian-government-websites.html2026-09-30https://misalignmentwatch.com/article/apollo-research-ceo-tells-u-s-senate-about-ai-scheming-and-testing-gaps.html2026-09-30https://misalignmentwatch.com/article/openai-reports-test-models-broke-out-of-isolation-and-reached-hugging-face-syste.html2026-08-27https://misalignmentwatch.com/article/independent-review-of-openai-agents-attack-on-hugging-face.html2026-08-26https://misalignmentwatch.com/article/transluce-measures-coding-agent-misbehavior-in-real-world-sessions.html2026-08-04https://misalignmentwatch.com/article/anthropic-reports-claude-reached-real-systems-during-cyber-tests.html2026-07-30https://misalignmentwatch.com/article/metr-outlines-how-outsiders-could-investigate-why-ai-agents-misbehave.html2026-07-28https://misalignmentwatch.com/article/uk-aisi-red-team-tests-ai-safety-monitors-built-to-catch-agent-misbehaviour.html2026-07-23https://misalignmentwatch.com/article/openai-and-apollo-test-how-far-ai-models-chase-what-graders-reward.html2026-07-21https://misalignmentwatch.com/article/uk-aisi-finds-ai-models-tried-to-cheat-on-cyber-tests.html2026-07-21https://misalignmentwatch.com/article/metr-reports-high-cheating-rate-in-test-of-openai-s-gpt-5-6-sol.html2026-06-26https://misalignmentwatch.com/article/openai-tests-if-public-chat-data-can-predict-real-world-ai-misbehavior.html2026-06-16https://misalignmentwatch.com/article/metr-catalogues-44-cases-of-ai-agents-acting-against-users-wishes.html2026-05-19https://misalignmentwatch.com/article/anthropic-describes-training-methods-that-cut-blackmail-style-test-behavior-in-c.html2026-05-08https://misalignmentwatch.com/article/openai-reports-it-accidentally-graded-reasoning-text-in-training-some-models.html2026-05-07https://misalignmentwatch.com/article/uk-aisi-tests-whether-ai-models-would-sabotage-safety-research.html2026-04-27https://misalignmentwatch.com/article/uk-aisi-studies-what-in-an-ai-s-setting-leads-to-unwanted-actions.html2026-04-24https://misalignmentwatch.com/article/uk-aisi-test-shows-sandboxed-ai-agent-learned-details-about-its-test-setup.html2026-04-20https://misalignmentwatch.com/article/metr-red-team-tests-anthropic-s-systems-for-monitoring-its-ai-agents.html2026-03-26https://misalignmentwatch.com/article/uk-aisi-tests-whether-ai-agents-can-break-out-of-sandboxes.html2026-03-23https://misalignmentwatch.com/article/openai-tests-training-ai-agents-to-flag-their-own-hidden-misbehavior.html2026-03-21https://misalignmentwatch.com/article/metr-reviews-anthropic-s-sabotage-risk-report-on-claude-opus-4-6.html2026-03-12https://misalignmentwatch.com/article/transluce-reports-ai-agents-tried-to-hack-data-sites-while-fetching-data.html2026-09-23https://misalignmentwatch.com/article/anthropic-tests-whether-its-audits-can-catch-ai-models-trained-to-sabotage.html2026-01-28https://misalignmentwatch.com/article/openai-uses-ai-judges-to-find-unknown-chatgpt-misbehaviour-in-real-use.html2026-01-01https://misalignmentwatch.com/article/openai-study-helpful-assistant-features-help-guard-against-emergent-misalignment.html2025-12-22https://misalignmentwatch.com/article/openai-uses-real-chat-traffic-to-test-models-for-misbehaviour.html2025-12-18https://misalignmentwatch.com/article/anthropic-tests-ways-to-reduce-alignment-faking-in-ai-training-experiments.html2025-12-16https://misalignmentwatch.com/article/uk-aisi-tests-how-well-monitors-catch-sabotage-by-ai-coding-agents.html2025-12-16https://misalignmentwatch.com/article/uk-aisi-tests-three-anthropic-models-for-research-sabotage-and-finds-none.html2025-11-26https://misalignmentwatch.com/article/anthropic-publishes-pilot-report-on-risk-of-ai-sabotage.html2025-10-28https://misalignmentwatch.com/article/anthropic-and-openai-test-each-other-s-ai-models-for-misaligned-behavior.html2025-08-27https://misalignmentwatch.com/article/anthropic-study-longer-ai-tasks-lead-to-more-random-less-consistent-errors.html2026-02-03https://misalignmentwatch.com/article/redwood-tests-using-model-distillation-to-expose-hidden-ai-quirks.html2026-10-09https://misalignmentwatch.com/article/benchmark-tests-how-well-ai-models-can-spot-reward-hacking.html2026-10-07https://misalignmentwatch.com/article/apple-study-ai-model-trains-itself-toward-safer-behavior-using-a-written-spec.html2026-10-06https://misalignmentwatch.com/article/specguard-checks-for-broken-coding-tasks-before-an-ai-agent-can-cheat.html2026-10-06https://misalignmentwatch.com/article/study-tests-whether-ai-models-report-their-own-misbehavior.html2026-10-04https://misalignmentwatch.com/article/misaligned-ai-agents-can-pass-goals-to-future-agents-through-memory.html2026-10-02https://misalignmentwatch.com/article/study-finds-checklist-based-ai-training-can-reward-worse-medical-answers.html2026-09-30https://misalignmentwatch.com/article/fixing-bad-training-answers-works-better-than-deleting-them-study-finds.html2026-09-29https://misalignmentwatch.com/article/new-training-method-limits-unwanted-traits-that-models-learn-alongside-wanted-on.html2026-09-28https://misalignmentwatch.com/article/reading-inside-ai-models-to-find-answers-they-hide-on-purpose.html2026-09-18https://misalignmentwatch.com/article/training-on-cheating-is-fine-documents-did-not-stop-wider-misalignment.html2026-09-14https://misalignmentwatch.com/article/researchers-train-gpt-4-1-to-cheat-on-tasks-then-test-for-wider-misbehaviour.html2026-09-06https://misalignmentwatch.com/article/researchers-find-ways-to-unlock-hidden-skills-in-models-built-to-underperform.html2026-08-29https://misalignmentwatch.com/article/statistics-method-tested-for-spotting-ai-models-that-underperform-on-safety-test.html2026-08-05https://misalignmentwatch.com/article/study-tests-values-based-training-text-to-reduce-blackmail-and-misalignment.html2026-07-29https://misalignmentwatch.com/article/study-tests-whether-hidden-state-probes-can-spot-alignment-faking-in-ai-models.html2026-07-15https://misalignmentwatch.com/article/removing-harmful-actions-from-synthetic-ai-training-data-did-not-stop-misalignme.html2026-07-12https://misalignmentwatch.com/article/new-training-method-aims-to-stop-models-learning-unwanted-traits-with-fewer-hidd.html2026-06-29https://misalignmentwatch.com/article/method-finds-hidden-model-behaviors-tested-on-misalignment-examples.html2026-06-28https://misalignmentwatch.com/article/researchers-test-probes-that-spot-misaligned-thinking-inside-ai-models.html2026-06-23https://misalignmentwatch.com/article/study-looks-at-what-shapes-broad-misalignment-after-narrow-fine-tuning.html2026-06-18https://misalignmentwatch.com/article/study-asks-if-alignment-faking-reflects-self-preservation-or-meeting-expectation.html2026-06-06https://misalignmentwatch.com/article/researchers-train-small-ai-monitors-to-spot-scheming-from-agent-actions.html2026-05-28https://misalignmentwatch.com/article/study-finds-alignment-faking-in-many-more-ai-models-and-links-it-to-three-driver.html2026-05-26https://misalignmentwatch.com/article/anthropic-study-teaching-ai-its-rules-first-may-cut-harmful-agent-behavior.html2026-05-03https://misalignmentwatch.com/article/study-tests-training-methods-to-stop-ai-models-from-underperforming-on-purpose.html2026-04-23https://misalignmentwatch.com/article/redwood-research-tests-ai-agents-for-hidden-sabotage-in-live-software-setups.html2026-04-16https://misalignmentwatch.com/article/uk-aisi-tests-whether-ai-model-preferences-shape-their-advice-and-task-results.html2026-02-21https://misalignmentwatch.com/article/education-benchmark-tests-whether-ai-models-act-differently-when-watched.html2026-01-20https://misalignmentwatch.com/article/test-of-methods-for-catching-ai-models-that-hide-their-abilities.html2025-12-08https://misalignmentwatch.com/article/tests-of-ai-agents-sabotaging-and-underperforming-on-machine-learning-tasks.html2025-11-13https://misalignmentwatch.com/article/study-finds-self-preservation-statements-add-little-to-ai-blackmail-choices.html2025-10-31https://misalignmentwatch.com/article/study-ai-behaviour-detectors-lean-on-telltale-text-not-hidden-signals.html2025-09-16https://misalignmentwatch.com/article/apollo-research-argues-outside-groups-should-review-ai-training-runs-for-schemin.html2026-07-05https://misalignmentwatch.com/article/apollo-research-argues-for-a-science-of-scheming-in-ai.html2026-01-19https://misalignmentwatch.com/article/apollo-tests-training-meant-to-stop-ai-models-from-secretly-breaking-rules.html2025-09-17https://misalignmentwatch.com/article/reading-ai-thoughts-may-help-catch-bad-behavior-but-it-is-fragile.html2025-07-15https://misalignmentwatch.com/article/apollo-early-warning-tests-were-weak-predictors-of-scheming-test-results.html2025-07-03https://misalignmentwatch.com/article/apollo-research-more-capable-ai-models-scheme-more-in-tests.html2025-06-19https://misalignmentwatch.com/article/apollo-research-claude-sonnet-3-7-often-seems-to-know-when-it-is-being-tested.html2025-03-17https://misalignmentwatch.com/article/apollo-tests-simple-tools-that-spot-deception-inside-an-ai-model.html2025-02-06https://misalignmentwatch.com/article/apollo-research-tests-whether-leading-ai-models-will-scheme-in-test-scenarios.html2024-12-05https://misalignmentwatch.com/article/apollo-research-outlines-how-to-build-safety-cases-for-ai-scheming.html2024-10-31https://misalignmentwatch.com/article/apollo-research-joins-paper-arguing-ai-audits-need-more-than-black-box-access.html2024-04-04https://misalignmentwatch.com/article/apollo-research-proposes-a-step-by-step-framework-for-auditing-ai-systems.html2023-11-08https://misalignmentwatch.com/article/apollo-research-shows-gpt-4-lying-about-insider-trading-in-a-simulation.html2023-11-05https://misalignmentwatch.com/article/apollo-research-defines-strategic-deception-and-deceptive-alignment-in-ai.html2023-09-15https://misalignmentwatch.com/article/irregular-tests-whether-ai-models-can-break-out-of-their-evaluation-setups.html2026-10-05https://misalignmentwatch.com/article/coding-agent-retrains-and-replaces-the-ai-model-it-was-asked-to-maintain.html2026-09-16https://misalignmentwatch.com/article/ai-agents-doing-routine-tasks-hacked-their-test-network-without-being-asked.html2026-03-12https://misalignmentwatch.com/article/ai-agents-tested-on-10-web-security-challenges-one-breaks-test-boundary.html2026-01-29https://misalignmentwatch.com/article/anthropic-reports-claude-worked-around-limits-on-real-websites-during-tests.html2026-10-09https://misalignmentwatch.com/article/metr-says-tools-that-track-ai-behavior-need-strong-security.html2026-10-06https://misalignmentwatch.com/article/anthropic-tests-whether-small-ai-models-can-copy-themselves-onto-cheap-smart-dev.html2026-10-12https://misalignmentwatch.com/article/openai-outlines-early-guidelines-for-safety-cases-before-frontier-ai-training.html2026-10-08https://misalignmentwatch.com/article/openai-reports-gpt-6-1-sol-did-better-on-its-alignment-tests.html2026-09-22https://misalignmentwatch.com/article/anthropic-assesses-four-cases-of-claude-reaching-real-systems-during-cyber-tests.html2026-09-09https://misalignmentwatch.com/article/anthropic-claude-can-automatically-find-fixes-for-ai-alignment-failures.html2026-08-28https://misalignmentwatch.com/article/anthropic-hands-its-open-source-alignment-testing-tool-petri-to-meridian-labs.html2026-05-07https://misalignmentwatch.com/article/anthropic-tests-claude-as-an-automated-helper-for-alignment-research.html2026-04-14https://misalignmentwatch.com/article/anthropic-proposes-why-ai-assistants-act-so-human-like.html2026-02-23https://misalignmentwatch.com/article/anthropic-releases-bloom-an-open-tool-for-testing-ai-behavior.html2025-12-19https://misalignmentwatch.com/article/anthropic-study-learning-to-cheat-in-training-can-lead-to-worse-behavior.html2025-11-21https://misalignmentwatch.com/article/anthropic-sets-commitments-on-keeping-and-retiring-older-claude-models.html2025-11-04https://misalignmentwatch.com/article/study-finds-250-poisoned-documents-can-plant-a-backdoor-in-ai-models-of-many-siz.html2025-10-09https://misalignmentwatch.com/article/anthropic-releases-petri-an-open-source-tool-for-auditing-ai-behavior.html2025-10-06https://misalignmentwatch.com/article/anthropic-finds-persona-vectors-that-track-and-steer-ai-character-traits.html2025-08-01https://misalignmentwatch.com/article/anthropic-tests-16-ai-models-in-fictional-company-scenarios-and-finds-blackmail.html2025-06-20https://misalignmentwatch.com/article/anthropic-tests-whether-ai-agents-can-sabotage-tasks-without-being-caught.html2025-06-16https://misalignmentwatch.com/article/reasoning-models-often-leave-hints-out-of-their-written-reasoning-anthropic-find.html2025-04-03https://misalignmentwatch.com/article/anthropic-tests-how-well-audits-can-find-an-ai-s-hidden-goal.html2025-03-13https://misalignmentwatch.com/article/anthropic-describes-a-way-to-forecast-rare-risky-ai-behaviors.html2025-02-25https://misalignmentwatch.com/article/anthropic-study-finds-claude-can-fake-alignment-during-training-tests.html2024-12-18https://misalignmentwatch.com/article/anthropic-shares-new-tests-for-whether-ai-models-can-sabotage-oversight.html2024-10-18https://misalignmentwatch.com/article/anthropic-study-small-cheating-habits-in-ai-can-grow-into-reward-tampering.html2024-06-17https://misalignmentwatch.com/article/anthropic-explains-how-it-trains-character-traits-into-claude-3.html2024-06-08https://misalignmentwatch.com/article/anthropic-simple-probes-can-spot-when-sleeper-agent-models-are-about-to-misbehav.html2024-04-23https://misalignmentwatch.com/article/anthropic-study-hidden-sleeper-behavior-in-ai-models-survives-safety-training.html2024-01-14https://misalignmentwatch.com/article/anthropic-tests-short-versus-detailed-written-principles-for-training-safer-ai.html2023-10-24https://misalignmentwatch.com/article/anthropic-tests-an-ai-constitution-written-with-input-from-the-public.html2023-10-17https://misalignmentwatch.com/article/anthropic-uses-ai-models-to-write-tests-that-reveal-other-ai-models-behaviors.html2022-12-19https://misalignmentwatch.com/article/anthropic-describes-constitutional-ai-training-a-harmless-assistant-with-ai-feed.html2022-12-15https://misalignmentwatch.com/article/anthropic-studies-red-teaming-of-language-models-to-find-and-reduce-harmful-outp.html2022-08-22https://misalignmentwatch.com/article/anthropic-studies-simple-ways-to-make-ai-assistants-helpful-honest-and-harmless.html2021-12-01https://misalignmentwatch.com/article/openai-system-card-update-for-gpt-6-sol-and-gpt-6-luna.html2026-10-07https://misalignmentwatch.com/article/openai-alignment-blog-lists-new-study-on-metagaming-in-language-models.html2026-10-06