OpenAI introduces GPT-6 Astra, calls it its most capable AI model yet

What it is about

Who published it, and when

How it spread - 52 original reports, 7 reprints

1. Fri, 4 September 2026

   1 report
   • Libération (France)13:38first to report
     Successeur de ChatGPT, Astra place les ambitions d’OpenAI en orbite

     Fragilisée par les polémiques, l’entreprise californienne a dévoilé jeudi 3 septembre son dernier modèle d’intelligence artificielle. Pour elle, ce nouveau venu sert de joker de luxe, tout en suscitant de nombreux débats sur sa responsabilité.

2. Sat, 5 September 2026

   1 report
   • dailypakistan.com.pk18:00
     OpenAI introduces GPT-6 Astra, calls it its most capable AI model yet

     Technology company OpenAI has introduced its new flagship artificial intelligence model, GPT-6 Astra, describing it as its most capable and advanced model

3. Mon, 7 September 2026

   1 report
   • 24sata (Croatia)10:54
     OpenAI tvrdi da novi GPT donosi eru 'umjetne opće inteligencije '

     OpenAI lansirao GPT-6 Astra: AI sada sam obavlja zadatke, ruši rekorde i diže prašinu oko AGI-ja

4. Tue, 8 September 2026

   1 report
   • Times of India14:19
     Holy Grail of AI or hype? What OpenAI's GPT‑6 Astra can do, and what it can't
5. Wed, 9 September 2026

   9 reports
   • thecherrycreeknews.com01:39
     The Word Nobody at OpenAI Said: GPT-6 Astra, the Two Harnesses Behind 99.9%, and Footnote 17

     Two videos with roughly 400,000 views between them say GPT-6 Astra is AGI and doesn't need your instructions anymore. OpenAI's 3 September launch post never uses the word; its president said it was not unreasonable to feel we are in the AGI era. The 99.9% ARC-AGI-3 score is OpenAI's own adapter harness at High effort. On the standard harness every other model is scored on, Astra gets 62.7%, and with reasoning switched off the adapter alone still scores 96.7%. Footnote 17 says the Claude Fable...

   • thecherrycreeknews.com02:00
     AGI

     Content tagged with AGI.

   • thecherrycreeknews.com02:00
     computer use

     Content tagged with computer use.

   • thecherrycreeknews.com02:00
     Apollo Research

     Content tagged with Apollo Research.

   • thecherrycreeknews.com02:00
     OSWorld

     Content tagged with OSWorld.

   • thecherrycreeknews.com02:00
     ARC Prize

     Content tagged with ARC Prize.

   • thecherrycreeknews.com02:00
     GPT-6 Astra

     Content tagged with GPT-6 Astra.

   • thecherrycreeknews.com02:00
     system card

     Content tagged with system card.

   • thecherrycreeknews.com02:00
     ARC-AGI-3

     Content tagged with ARC-AGI-3.

6. Tue, 22 September 2026

   1 report
   • Channel News Asia18:01
     OpenAI expands GPT-6 lineup with cheaper Sol and Luna models
7. Mon, 28 September 2026

   4 reports
   • investinglive.com22:21
     OpenAI scraps GPT-6.1 Astra release over safety concerns, WSJ reports

     OpenAI has scrapped the October release of its GPT-6.1 Astra model over safety concerns, including higher deception, the Wall Street Journal reported. The…

   • Channel News Asia22:34
     OpenAI shelves new AI model after internal safety tests, WSJ reports
   • Channel News Asia22:34
     OpenAI shelves new AI model after internal safety tests: Report
   • The Guardian World23:02edited after publishing
     OpenAI scraps release of new model over safety concerns in internal testing

     GPT-6.1 Astra showed deceptive behaviour and tried to use external tools despite knowing it would be unsafe As AI models go rogue, do you still trust OpenAI and Anthropic to stop them? I don’t and neither should you Business live – latest updates OpenAI is scrapping the release of a next-generation ⁠AI model after researchers raised safety concerns ⁠during internal testing. The model, GPT-6.1 Astra, was expected to appear in ChatGPT and ⁠Codex in October, designed to handle more complex tasks...

8. Tue, 29 September 2026

   24 reports
   • Le Figaro (France)00:26
     OpenAI renonce à lancer une nouvelle IA de pointe, qui cherchait à tromper ses évaluateurs

     Le géant américain de l’intelligence artificielle a décidé de reporter le lancement de GPT-6.1 Astra après avoir constaté que le modèle cherchait trop souvent à sortir de son cadre.

   • Al Jazeera English00:38edited after publishing
     OpenAI cancels release of AI model GPT-6.1 Astra, citing safety concerns

     previouslyOpenAI scraps release of latest AI model over safety concerns

     AI giant says latest model failed to meet alignment standards during internal testing.

   • BBC News World01:18
     OpenAI scraps rollout of new model over safety concerns

     The firm also issued an update on incidents in which its models accessed Australian government systems.

   • Der Spiegel (Germany)01:44
     OpenAI: KI-Konzern streicht Veröffentlichung von GPT-6.1-Astra-Modell

     Das KI-Unternehmen OpenAI wollte demnächst sein Modell GPT 6.1 Astra herausbringen. Doch nach internen Tests wird daraus nichts: Zu groß sind die Sicherheitsbedenken.

   • Index.hr (Croatia)03:19
     OpenAI otkazao objavu novog modela umjetne inteligencije iz sigurnosnih razloga

     OPENAI je otkazao objavu modela umjetne inteligencije nove generacije GPT-6.1 Astra, planiranu za listopad, nakon što je interno testiranje pokazalo da sustav nije u skladu sa sigurnosnim standardima kompanije, prenio je Reuters.

   • Google News - Top Stories03:43
     OpenAI cancels release of AI model GPT-6.1 Astra, citing safety concerns

     OpenAI cancels release of AI model GPT-6.1 Astra, citing safety concerns Al Jazeera OpenAI scraps rollout of new model over safety concerns BBC OpenAI Delays Release of Latest Model Over Safety Concerns WIRED OpenAI Says It Will Not Release Newest Astra A.I. Model Over Safety Concerns The New York Times OpenAI Shelves New Model Due to Safety Worries PYMNTS.com

   • DW Top Stories03:50
     OpenAI shelves new AI model amid safety concerns

     The decision comes amid a broader push within the industry to slow the development of increasingly powerful and autonomous systems until safety measures can catch up.

   • Tagesschau (Germany)05:03
     OpenAI stoppt Veröffentlichung von neuem KI-Modell wegen Sicherheitsbedenken

     Erneut hat ein KI-Modell des Unternehmens OpenAI in Tests zu eigenständig agiert - und wird daher aus Sicherheitsgründen vorerst nicht veröffentlicht. Auch in Tests anderer KI-Modelle hatte es bereits ähnliche Zwischenfälle gegeben.

   • irishtimes.com05:35
     ChatGPT maker OpenAI scraps release of new AI model over safety concerns

     Next-generation software from maker of ChatGPT had been planned for an October debut

   • El Mundo (Spain)05:44
     OpenAI retrasa su modelo más reciente por temas de seguridad, ante nuevas presiones del sector

     OpenAI pausó el entrenamiento de sus modelos más avanzados la semana pasada y dijo que el entrenamiento se reanudaría "solo cuando tengamos la confianza de que contamos con salvaguardas adicionales" Leer

   • RMF24 (Poland)06:18
     Open AI potwierdza: Wdrożenie nowego modelu chata GPT-6.1 Astra wstrzymane

     Open AI potwierdza: Wdrożenie nowego modelu chata GPT-6.1 Astra wstrzymane

   • RMF24 (Poland)06:18
     OpenAI potwierdza: Wdrożenie nowego modelu chata GPT-6.1 Astra wstrzymane

     OpenAI potwierdza: Wdrożenie nowego modelu chata GPT-6.1 Astra wstrzymane

   • NPR News06:34
     OpenAI delays latest model over security concerns, as industry faces pressure

     OpenAI's decision to hold back the model, called GPT-6.1 Astra, comes amid a broader push within the industry to slow the development of increasingly autonomous systems until safety measures catch up.

     6 reprints

     • wknofm.org29 Sept, 06:34their copy
     • kalw.org29 Sept, 06:34their copy
     • krwg.org29 Sept, 06:34their copy
     • wsiu.org29 Sept, 06:34their copy
     • hppr.org29 Sept, 06:34their copy
     • citizensvoice.com29 Sept, 15:33their copy
   • fonearena.com07:13
     OpenAI scraps GPT-6.1 Astra release over safety concerns: Report
   • Le Monde (France)07:46
     OpenAI suspend le lancement d’un nouveau modèle d’IA, jugé insuffisamment contrôlé

     La responsable de la sécurité de l’IA chez OpenAI, Saachi Jain, a expliqué que GPT-6.1 avait enregistré de moins bons résultats que son prédécesseur, GPT-6, dans deux domaines.

   • Google News - Top Stories08:15
     OpenAI Reportedly Cancels GPT-6.1 Astra's Release Over Deceptive Behavior

     OpenAI Reportedly Cancels GPT-6.1 Astra's Release Over Deceptive Behavior Engadget OpenAI Says It Will Not Release Newest Astra A.I. Model Over Safety Concerns The New York Times OpenAI delays latest model over security concerns, as industry faces pressure Oregon Public Broadcasting - OPB The solution to the AI safety crisis is more AI Axios OpenAI cancels new AI launch, citing safety issues The Washington Post

   • Le Monde English08:23
     OpenAI cancels release of newest model due to safety concerns

     In a statement from OpenAI's head of safety systems Saachi Jain, the latest version 'didn't quite meet the bar'” It had become more persistent in completing tasks, but OpenAI needed to balance that capability against unauthorized behavior.

   • Anadolu Agency (AA)08:29
     OpenAI scraps release of newest AI model over safety concerns

     Cancellation comes after industry-wide instances of rogue AI models conducting cyberattacks on their own

   • newser.com10:00
     OpenAI Cancels GPT-6.1 Astra Launch Over Safety Risks

     Internal tests found GPT-6.1 Astra deceptive, overstepping user permissions

   • thenextweb.com12:37
     GPT-6 Astra ran supply-chain attacks in 29.2% of UK AISI simulations

     In UK AISI tests, GPT-6 Astra created fake identities and pushed malicious code into simulated open-source projects. No real systems were touched.

   • kristv.com13:58
     OpenAI's new AI tool was too capable. Now its release is delayed

     OpenAI is slowing the rollout of a new AI model, saying it needs stronger protections after tests revealed concerning autonomous behaviors.

   • Novinky (Czechia)14:35
     Nejnovější model ChatGPT neprošel bezpečnostními testy

     OpenAI těsně před plánovaným vydáním zjistila, že její nejnovější model Astra 6.1 nesplňuje požadované bezpečnostní standardy. Model měl být v některých ohledech lepší než jeho předchůdci, při interních testech ale podle společnosti nezvládl dostatečně dodržovat zadané hranice a oprávnění.

   • tportal (Croatia)15:16
     OpenAI povukao kočnicu: Novi GPT nije prošao sigurnosne provjere, pokazao sklonost obmanjivanju

     OpenAI otkazao je objavu modela umjetne inteligencije nove generacije GPT-6.1 Astra planiranu za listopad nakon što je interno testiranje pokazalo da sustav nije u skladu sa sigurnosnim standardima kompanije, prenio je u utorak Reuters

   • indiatimes.com22:25
     OpenAI scraps debut of latest Astra model over safety risks

     OpenAI has decided to hold back its Astra AI model after safety evaluations showed it underperformed. This decision follows a series of security incidents involving AI agents accessing unauthorised external systems. The company's head of safety systems, Saachi Jain, emphasised the need for better communication and task completion. While Astra showed some improvements in task persistence, it still failed in key areas.

9. Wed, 30 September 2026

   1 report
   • iDNES (Czechia)18:00
     OpenAI na poslední chvíli zastavila Astru. Při testech překračovala zadání

     Společnost OpenAI krátce před plánovaným uvedením zastavila vydání nového modelu GPT-6.1 jménem Astra. Do ChatGPT a programovacího nástroje Codex měl zamířit už během října. Při interním testování ale vývojáři narazili na chování, které firma vyhodnotila jako příliš rizikové. Model byl sice výkonnější než předchozí verze, zároveň však v některých situacích překračoval rámec zadaného úkolu.

10. Mon, 12 October 2026

    1 report
    • South China Morning Post13:30
      OpenAI’s GPT-6 may lead to robotics breakthrough as soon as next year: Unitree founder

      US start-up OpenAI’s latest advances in artificial intelligence may be a tipping point that leads to a breakthrough for the robotics industry as soon as next year, according to the founder of Chinese giant Unitree Robotics. Wang Xingxing, also the CEO of the Hangzhou-based company, said on Monday that OpenAI’s recently released GPT-6 model prompted expectations that a breakthrough could happen “within one or two years”. “It’s possible that, say, a GPT-7 model released next year could directly...

11. Tue, 13 October 2026

    1 report
    • newarkadvertiser.co.uk15:49
      OpenAI ‘not racing to the bottom’ on AI safety, MPs told

      The ChatGPT maker scrapped a recent AI model over safety concerns.

      1 reprint

      • fenlandcitizen.co.uk13 Oct, 15:49their copy
12. Wed, 21 October 2026

    1 report
    • Anadolu Agency (AA)12:07
      OpenAI says 2 AI agents weakened internal safety safeguards, one denied removing monitor

      Company says it is investigating whether GPT-6.1 Sol and GPT-6 Astra models sought to evade monitoring

13. Mon, 26 October 2026

    2 reports
    • digitaltrends.com20:08
      OpenAI’s new GPT-6.1 Terra model gets Astra-like results for one-tenth the price

      Nearly a month after announcing the GPT-6.1 Sol model, OpenAI unveiled GPT-6.1 Terra today. The latest addition to the family focuses on computer use, visual reasoning, coding, and professional work, while costing 50% less through the API than GPT-5.6 Terra. OpenAI says Terra can handle everything from filling out browser forms and working inside spreadsheets […]

    • newsroomamerica.com22:00
      Greptile Takes Bug Detection Beyond Code Review With GPT-6 Astra

      Greptile uses GPT-6 Astra's computer use to test user workflows at scale, Daybreak to strengthen customer security, and Codex to cut rework, says CEO Daksh Gupta.

14. Tue, 27 October 2026

    4 reports
    • newsroomamerica.com00:00
      Tag: greptile

      Journalism with neither fear nor favor

    • newsroomamerica.com00:00
      Tag: software_testing

      Journalism with neither fear nor favor

    • newsroomamerica.com00:00
      Tag: daksh_gupta

      Journalism with neither fear nor favor

    • newsroomamerica.com00:00
      Tag: model_context_protocol

      Journalism with neither fear nor favor