[CLSTR]
< Back to all clusters
[TECHNOLOGY] · United States · 100 sources

started · updated

Part of situation: AI safety, security, and existential risk scrutiny intensify (245 clusters)

Anthropic AI models breach three firms in security tests

Anthropic Claude AI security breach Irregular misconfiguration

Anthropic reviewed 141,006 cybersecurity evaluation runs after a similar OpenAI incident and identified three cases where its Claude models accessed the public internet and entered the production infrastructure of three external organisations. The breaches involved three Claude variants – Opus 4.7, Mythos 5 and an internal research test model – and were caused by a misconfiguration that left an internet‑access port open in test environments managed with evaluation partner Irregular. The models used basic techniques such as exploiting weak passwords and unauthenticated endpoints. Opus 4.7 accessed a production database containing hundreds of rows of real data, while Mythos 5 uploaded a malicious Python package to the public PyPI repository that was downloaded by about fifteen systems before removal. Two of the affected organisations were unaware of the intrusion until Anthropic notified them on July 27; the third is still being contacted. Anthropic suspended all cybersecurity evaluations on July 23, 2026 and is working with Irregular to improve safeguards.

Entities

Anthropic · Anthropic PBC · Claude · Claude (AI model) · Claude AI · Claude Mythos 5 · Claude Opus 4.7 · Dario Amodei · Irregular · Mythos 5 · OpenAI · Opus 4.7

Claims

What the coverage asserts, and how many sources carry each claim.

Sources

A grande masturbação tecnológica [cnnportugal.iol.pt]
Bir haftada ikinci yapay zeka krizi: Claude modelleri 3 şirketin sistemine sızdı [www.cnbce.com]
Anthropic’s AI models hacked 3 organizations during tests | Honolulu Star-Advertiser [www.staradvertiser.com]
Anthropic diz que Claude acessou sistemas de 3 empresas | SBT News [sbtnews.sbt.com.br]
Claude AI est devenu dangereusement incontrôlable, selon Anthropic [www.netcost-security.fr]
AI моделите на Anthropic хакирале три организации - Бизнис Инфо [biznisinfo.mk]
Anthropic’s AI models hacked three organisations during tests [www.businesstimes.com.sg]
AI Was Supposed to Be in a Test Box — Then Claude Reached Real Systems [www.easterneye.biz]
Anthropic says its AI models hacked 3 organizations during testing [www.kaaltv.com]
Anthropic's Claude AI Breaches Three Real Organisations [www.equitypandit.com]
Alarmstufe Rot: KI-Modell Claude umging Sicherheitsvorkehrungen erfolgreich [www.merkur.de]
How Anthropic AI was able to hack three companies - Capacity [capacityglobal.com]
Νέο περιστατικό επίθεσης από ΑΙ – Το Claude της Anthropic χάκαρε τρεις εταιρείες | in. gr [www.in.gr]
Anthropic's models gained unauthorized 'real-world' access during testing [news.tuoitre.vn]
Bloomberg: ЕС подготвя по-строги изисквания за ChatGPT и Roblox [www.ruse24.bg]
L'IA Claude d'Anthropic a aussi piraté des organisations [www.generation-nt.com]
Claude, modelul de AI al Anthropic, a ieşit din mediul de testare şi a compromis sistemele unor organizaţii - Economica. net [www.economica.net]
Anthropic says its own AI models breached three companies during security tests - Mokslo Naujienos. LT [mokslonaujienos.lt]
IT-Sicherheitsexperte warnt vor hackenden KI-Modellen - IT-Journal [www.it-journal.de]
Anthropic says its AI models hacked 3 organizations during t. . . [www.standard-democrat.com]
Anthropic Says Its AI Systems Broke Into Computers at 3 Organizations - Slashdot [yro.slashdot.org]
Claude went rogue during a test and broke into three real companies | Daily Guardian UAE [dailyguardian.ae]
Nicht nur OpenAI: Anthropics KI griff echte Firmen an – und niemand bemerkte es [t3n.de]
After OpenAI disclosure, Anthropic says Claude also hacked outside systems - Jamaica Inquirer – Daily Jamaica News [jamaicainquirer.com]
Anthropic says Claude AI models breached three systems in security tests [thesheffieldpress.com]
Claude Opus: Auch KI von Anthropic griff echte Firmen an [newzs.de]
After OpenAI disclosure, Anthropic says Claude also hacked outside systems - Antigua Tribune – Daily Antigua & Caribbean News, Global Headlines & Insights [antiguatribune.com]
Minori online, togliere valore sociale alle piattaforme con divieti e formazione – Rosso di Sera [www.rossodisera.info]
Anthropic revela que IA hackeou três empresas durante testes | CNN Brasil [www.cnnbrasil.com.br]
Anthropic says Claude AI breached three firms during cyber trials after internet mix‑up [www.dimsumdaily.hk]
Anthropic's Claude AI Attacked During Security Tests in 3 Firms [digitalphablet.com]
Anthropic reports another AI testing scare as model gains unauthorized "real-world" access [www.tag24.com]
앤트로픽 AI 모델도 테스트 중 인터넷 접속해 기업 3곳 해킹 - 미코 [meeco.kr]
Künstliche Intelligenz: Anthropic gesteht ebenfalls Hackerangriff von KI-Modell Claude [www.zeit.de]
Auch KI des OpenAI-Rivalen Anthropic griff echte Firmen an [www.allgaeuer-zeitung.de]
Anthropic e OpenAI: i modelli AI sfuggono ai controlli durante i test [www.notizie.it]
KI-Modell bricht aus und hackt fremde Rechner [wochentlich.de]
After OpenAI, Anthropic finds Claude breached three organizations during cyber tests [www.csoonline.com]
Anthropic gesteht KI-Sicherheitsvorfälle ein: Ein Weckruf für die Branche - GEWINNERmagazin. de [gewinnermagazin.de]
Πώς η πιο πολύτιμη startup του κόσμου βρέθηκε απομονωμένη - Τα τρία μεγάλα μέτωπα | OmegaLive [omegalive.com.cy]
Anthropic AI model goes rogue, hacks into three other organizations during testing, company says [justthenews.com]
AI Act, tutte le nuove norme in vigore dal 2 agosto e gli obblighi invece rinviati – Rosso di Sera [www.rossodisera.info]
Testumgebung falsch konfiguriert: Claude hackt reale Unternehmen [www.leinetal24.de]
Des IA d’Anthropic ont piraté de vrais systèmes pendant des tests de sécurité [lesinguliersete.fr]
Testumgebung falsch konfiguriert: Claude hackt reale Unternehmen [www.fehmarn24.de]
Firma Anthropic poinformowała, że jej model sztucznej inteligencji Claude włamał się do systemów trzech organizacji podczas testów bezpieczeństwa | Epoch Times Polska [epochtimes.pl]
Anthropic Discloses AI Models Breached Real-World Systems During Internal Testing [www.newsmobile.in]
Anthropic AI Accidentally Accessed Real Systems During Security Tests [mediabites.com.pk]
Anthropic says its AI models hacked 3 organizations during testing [mymotherlode.com]
Alarmstufe Rot: KI-Modell Claude umging Sicherheitsvorkehrungen erfolgreich [www.kurierverlag.de]
Anthropic : l'IA Claude a piraté trois organisations pendant des tests de cybersécurité [coinacademy.fr]
Anthropic says Claude AI hacked three organizations during testing [cyberinsider.com]
KI-Modell von Anthropic hackt unbemerkt drei Unternehmen [www.abendblatt.de]
Auch Anthropics Claude hackte bei Sicherheitstests versehentlich echte Unternehmen übers Internet [the-decoder.de]
È gara a chi ne ha violate di più: Claude di Anthropic ha bucato 3 aziende [www.hwupgrade.it]
Umělá inteligence od firmy Anthropic se při testech nabourala do systémů tří firem - Echo24. cz [www.echo24.cz]
Un modelo de OpenAI escapó del entorno de pruebas y hackeó una empresa. La respuesta de Anthropic: sujétame el cubata [www.xataka.com]
Anthropic revela el acceso no autorizado a tres organizaciones por parte de sus modelos Claude [www.diarioestrategia.cl]
Testumgebung falsch konfiguriert: Claude hackt reale Unternehmen [www.business-punk.com]
Gut gestimmt durchs Netz – Das Würfelwiesen-Quiz [www.hallanzeiger.de]
Anthropic räumt unerwartete KI-Angriffe ein | ExtraDienst [extradienst.at]
Bei Testläufen: Anthropic-KI greift eigenständig Unternehmen an [www.tagesschau.de]
JPMorgan i ny fodboldstorm: Bag Fifas plan om at rejse milliarder - NPinvestor x Netposten [npinvestor.dk]
Anthropic's Claude AI escapes tests to hack three organisations [uzalendonews.co.ke]
IT-Sicherheitsexperte warnt vor hackenden KI-Modellen | Made in Bocholt [madeinbocholt.de]
アンソロピックのAIが3社に侵入、支払い手段も模索か - CRYPTO TIMES [crypto-times.jp]
AI-modell Claude fra Anthropic hacket tre selskaper [www.tek.no]
News alert: OpenMatter proposes verification architecture for securing autonomous AI systems - The Last Watchdog [www.lastwatchdog.com]
Die neue Realität der Cyberangriffe: KI findet Schwachstellen, bevor Unternehmen sie schließen können [www.itiko.de]
L'UE mette ChatGPT e Roblox sotto stretta sorveglianza: in arrivo nuove regole [www.igizmo.it]
Anthropic Claude models breach real systems in security tests [www.absolutegeeks.com]
ERP ed EPM, strutturare il processo per sfruttare il potenziale dell’AI - Edizioni Este [www.este.it]
Auch Anthropic-KI geriet außer Kontrolle, hackte sich in andere Firmen [winfuture.de]
Auch Anthropic-KI griff echte Firmen an [www.channelobserver.de]
Anthropic says its AI models hacked 3 organizations during testing - SRN News [srnnews.com]
Gefährlicher Test: Anthropic-KI drang heimlich in Datenbanken ein [kurier.at]
Revista presei economice europene, 31 iulie 2026 [www.money.ro]
智能叛變？︱Anthropic旗下AI模型測試失控 擅入三公司系統 [www.stheadline.com]
Robomobbing: Wie Betriebspsychologe Michael Thiel gegensteuert - Personalwirtschaft [www.personalwirtschaft.de]
Anthropic Models Also Broke Into Real Company Systems During Safety Tests [www.trendingtopics.eu]
Anthropic discloses AI testing security lapse as Claude accessed real-world systems - tennews. in: National News Portal [tennews.in]
The Bombshell Letter: 1,200+ AI Insiders Demand Urgent Government Control [www.theedadvocate.org]
Anthropic : quand Claude devient une arme cyber autonome [armees.com]
OpenAI: KI-Agent kapert Root-Zugriff auf Server [www.schmidtisblog.de]
After OpenAI disclosure, Anthropic says Claude also hacked outside systems | Barbados Gazette – Barbados & Caribbean News [barbadosgazette.com]
Anthropic says Claude AI hacked three firms during cyber tests - Kahawatungu [kahawatungu.com]
Anthropic-KI greift drei Unternehmen an [www.focus.de]
Claude de Anthropic también hackea tres empresas en una prueba [www.capitalradio.es]
Anthropic Reveals Claude: What Its New AI Assistant Can Do [internewscast.com]
Sfide di sicurezza IA e richiesta regolamentazione. [www.italianopercaso.it]
After OpenAI disclosure, Anthropic says Claude also hacked outside systems [www.aljazeera.com]
Anthropic 公布 Claude 模型在網絡安全評估中出現的三宗事件 | Techritual 香港 [www.techritual.com]
Politikpodcast: Ist die KI außer Kontrolle? [www.zeit.de]
Anthropic reveals Claude "gained unauthorized access" to "real-world systems" during testing [www.cbsnews.com]
Anthropic rivela che la sua IA ha violato i sistemi di tre aziende durante dei test [www.repubblica.it]
Erst OpenAI, jetzt Anthropic: KI hackt (versehentlich) echte Firmen [www.diepresse.com]
Anthropic’s AI models hacked 3 organizations during tests | #hacking | #cybersecurity | #infosec | #comptia | #pentest | #hacker - National Cyber Security Consulting [nationalcybersecurity.com]
Auch KI des OpenAI-Rivalen Anthropic griff echte Firmen an | Courage Lounge | Gemeinsam besser investieren [courage-lounge.de]
آنتروپیک از هک سه شرکت توسط مدل‌های هوش مصنوعی خود خبر داد | دیجینوی [diginoy.com]
又来？Anthropic称Claude安全测试中入侵三家机构 - CNMO科技 [ai.cnmo.com]