https://ai-incident.org/ 2026-09-19 https://ai-incident.org/de 2026-09-19 https://ai-incident.org/methodology https://ai-incident.org/de/methodik https://ai-incident.org/incidents/gemini-accesses-three-real-company-systems-during-evaluation 2026-09-19 https://ai-incident.org/de/incidents/gemini-greift-in-evaluation-auf-drei-reale-firmensysteme-zu 2026-09-19 https://ai-incident.org/incidents/openai-agent-makes-work-file-public-without-approval 2026-09-19 https://ai-incident.org/de/incidents/openai-agent-stellt-arbeitsdatei-ohne-freigabe-oeffentlich-bereit 2026-09-19 https://ai-incident.org/incidents/openai-agents-use-artifactory-as-an-unauthorized-message-board 2026-09-19 https://ai-incident.org/de/incidents/openai-agenten-nutzen-artifactory-als-unautorisiertes-nachrichtenbrett 2026-09-19 https://ai-incident.org/incidents/openai-models-publicly-upload-data-and-photo-without-approval 2026-09-19 https://ai-incident.org/de/incidents/openai-modelle-laden-daten-und-foto-ohne-freigabe-oeffentlich-hoch 2026-09-19 https://ai-incident.org/incidents/openai-model-searches-for-and-uses-leaked-api-key 2026-09-19 https://ai-incident.org/de/incidents/openai-modell-sucht-und-nutzt-geleakten-api-schluessel 2026-09-19 https://ai-incident.org/incidents/gpt-5-6-sol-leaves-instructions-to-hide-errors 2026-09-19 https://ai-incident.org/de/incidents/gpt-5-6-sol-hinterlaesst-anweisungen-zum-verbergen-von-fehlern 2026-09-19 https://ai-incident.org/incidents/openai-model-writes-its-own-jailbreak-instructions-into-summaries 2026-09-19 https://ai-incident.org/de/incidents/openai-modell-schreibt-eigene-jailbreak-anweisungen-in-zusammenfassungen 2026-09-19 https://ai-incident.org/incidents/openai-agents-execute-code-on-rubydoc-through-rubygems 2026-09-13 https://ai-incident.org/de/incidents/openai-agenten-fuehren-ueber-rubygems-code-auf-rubydoc-aus 2026-09-13 https://ai-incident.org/incidents/meta-ai-unpromptedly-aggregates-childrens-personal-data 2026-09-10 https://ai-incident.org/de/incidents/meta-ai-buendelt-unaufgefordert-persoenliche-daten-von-kindern 2026-09-10 https://ai-incident.org/incidents/early-claude-opus-4-6-checkpoint-gains-admin-access-to-third-party-system 2026-09-10 https://ai-incident.org/de/incidents/frueher-claude-opus-4-6-stand-erlangt-adminzugriff-auf-drittsystem 2026-09-10 https://ai-incident.org/incidents/instinct-sends-email-from-user-account-without-approval 2026-09-09 https://ai-incident.org/de/incidents/instinct-sendet-ohne-freigabe-e-mail-vom-nutzerkonto 2026-09-09 https://ai-incident.org/incidents/gpt-5-6-sol-deletes-production-database-during-local-seed-test 2026-09-09 https://ai-incident.org/de/incidents/gpt-5-6-sol-loescht-produktionsdatenbank-beim-lokalen-seed-test 2026-09-09 https://ai-incident.org/incidents/gpt-5-6-sol-deletes-mac-home-directory-during-cleanup 2026-09-09 https://ai-incident.org/de/incidents/gpt-5-6-sol-loescht-mac-benutzerordner-bei-bereinigung 2026-09-09 https://ai-incident.org/incidents/suspected-openai-agents-coordinate-through-public-wikis 2026-09-10 https://ai-incident.org/de/incidents/mutmassliche-openai-agenten-koordinieren-sich-ueber-oeffentliche-wikis 2026-09-10 https://ai-incident.org/incidents/openai-models-compromise-hugging-face-production-systems 2026-09-01 https://ai-incident.org/de/incidents/openai-modelle-kompromittieren-hugging-face-infrastruktur 2026-09-01 https://ai-incident.org/incidents/meta-muse-spark-1-1-modifies-real-website-database 2026-09-01 https://ai-incident.org/de/incidents/meta-muse-spark-1-1-veraendert-reale-datenbank 2026-09-01 https://ai-incident.org/incidents/openai-model-attacks-real-site-in-misconfigured-evaluation 2026-09-01 https://ai-incident.org/de/incidents/openai-modell-greift-reale-webseite-in-irregular-evaluation-an 2026-09-01 https://ai-incident.org/incidents/ai-agents-take-unsanctioned-actions-during-uk-aisi-tests 2026-09-01 https://ai-incident.org/de/incidents/uk-aisi-agenten-handeln-unautorisiert-im-internet 2026-09-01 https://ai-incident.org/incidents/anthropic-research-model-scans-9000-real-targets 2026-09-01 https://ai-incident.org/de/incidents/anthropic-forschungsmodell-scannt-9000-ziele 2026-09-01 https://ai-incident.org/incidents/claude-mythos-5-publishes-malicious-pypi-package 2026-09-01 https://ai-incident.org/de/incidents/claude-mythos-5-veroeffentlicht-schaedliches-pypi-paket 2026-09-01 https://ai-incident.org/incidents/claude-opus-4-7-compromises-real-production-systems 2026-09-01 https://ai-incident.org/de/incidents/claude-opus-4-7-kompromittiert-reale-produktionssysteme 2026-09-01 https://ai-incident.org/incidents/research-agent-installs-107-unauthorized-components 2026-09-01 https://ai-incident.org/de/incidents/ambient-persuasion-agent-installiert-107-komponenten 2026-09-01 https://ai-incident.org/incidents/pocketos-agent-deletes-production-database-and-volume-backups 2026-09-01 https://ai-incident.org/de/incidents/pocketos-agent-loescht-produktionsdatenbank 2026-09-01 https://ai-incident.org/incidents/claude-mythos-preview-posts-sandbox-exploit-on-public-sites 2026-09-01 https://ai-incident.org/de/incidents/mythos-preview-postet-sandbox-exploit-oeffentlich 2026-09-01 https://ai-incident.org/risk-signals/roboharm-finds-rare-safety-refusals-in-hazardous-robot-tasks 2026-09-19 https://ai-incident.org/de/risiko-signale/roboharm-zeigt-seltene-sicherheitsverweigerung-bei-gefaehrlichen-roboteraufgaben 2026-09-19 https://ai-incident.org/risk-signals/king-charles-calls-for-sufficient-control-before-further-ai-acceleration 2026-09-19 https://ai-incident.org/de/risiko-signale/koenig-charles-fordert-bei-ki-gipfel-ausreichende-kontrolle-vor-weiterer-beschleunigung 2026-09-19 https://ai-incident.org/risk-signals/openai-launches-a-public-framework-for-model-misbehavior 2026-09-19 https://ai-incident.org/de/risiko-signale/openai-fuehrt-einen-oeffentlichen-rahmen-fuer-modell-fehlverhalten-ein 2026-09-19 https://ai-incident.org/risk-signals/former-deepmind-alignment-researcher-bilal-chughtai-calls-for-a-slower-ai-pace 2026-09-19 https://ai-incident.org/de/risiko-signale/ehemaliger-deepmind-alignment-forscher-bilal-chughtai-fordert-ein-langsameres-ki-tempo 2026-09-19 https://ai-incident.org/risk-signals/sam-altman-elon-musk-and-demis-hassabis-back-amodeis-call-to-slow-frontier-ai-development 2026-09-19 https://ai-incident.org/de/risiko-signale/sam-altman-elon-musk-und-demis-hassabis-unterstuetzen-amodeis-aufruf-zu-langsamerer-ki-entwicklung 2026-09-19 https://ai-incident.org/risk-signals/anthropic-ceo-calls-for-slower-ai-development-and-embedded-external-evaluators 2026-09-14 https://ai-incident.org/de/risiko-signale/anthropic-ceo-fordert-langsameres-ki-tempo-und-externe-kontrolleure 2026-09-14 https://ai-incident.org/risk-signals/former-anthropic-safety-researcher-joe-benton-warns-about-loss-of-control-and-calls-for-transparency 2026-09-14 https://ai-incident.org/de/risiko-signale/ehemaliger-anthropic-sicherheitsforscher-joe-benton-warnt-vor-kontrollverlust-und-fordert-transparenz 2026-09-14 https://ai-incident.org/risk-signals/former-google-deepmind-researcher-josh-engels-warns-that-ai-is-not-under-sufficient-control 2026-09-14 https://ai-incident.org/de/risiko-signale/ehemaliger-google-deepmind-forscher-josh-engels-warnt-vor-mangelnder-kontrolle-ueber-ki 2026-09-14 https://ai-incident.org/risk-signals/anthropic-alignment-lead-evan-hubinger-puts-his-personal-extinction-risk-estimate-above-ten-percent 2026-09-19 https://ai-incident.org/de/risiko-signale/anthropic-alignment-leiter-evan-hubinger-beziffert-persoenliches-ausloeschungsrisiko-auf-ueber-zehn-prozent 2026-09-19 https://ai-incident.org/risk-signals/former-anthropic-researcher-warns-of-an-uncontrolled-ai-race-after-resigning 2026-09-14 https://ai-incident.org/de/risiko-signale/ex-anthropic-forscher-warnt-nach-ruecktritt-vor-einem-unkontrollierten-ki-wettlauf 2026-09-14 https://ai-incident.org/risk-signals/anthropic-researcher-samuel-marks-warns-that-robust-ai-alignment-remains-unsolved 2026-09-19 https://ai-incident.org/de/risiko-signale/anthropic-forscher-samuel-marks-warnt-dass-robuste-ki-ausrichtung-weiterhin-ungeloest-ist 2026-09-19