[TECHNOLOGY] · United States · 100 sources

started 12 days ago · updated 10 days ago

Part of situation: OpenAI launches GPT-6 Astra amid AGI debate and security (16 clusters)

OpenAI launches GPT-6 Astra with critical cybersecurity capabilities

GPT-6 Astra cybersecurity autonomous agents AGI zero-day exploits

OpenAI has launched GPT-6 Astra, a new AI model that the company describes as its most intelligent and aligned to date. The model represents a significant shift toward autonomous AI agents, capable of performing complex, multi-step tasks such as coding, scientific research, and direct computer and browser usage. OpenAI President Greg Brockman suggested the release marks the beginning of the 'AGI era.'

Astra has become the first OpenAI model to reach the 'Critical' cybersecurity capability threshold under the company’s Preparedness Framework. This classification means the model can independently discover previously unknown zero-day vulnerabilities and develop functional exploits against hardened, real-world systems without human intervention. In internal testing, Astra achieved a 100% score on ExploitBench and demonstrated the ability to escape browser sandboxes and escalate privileges to root-level access.

Due to these high-level offensive capabilities, OpenAI is implementing strict safeguards. Advanced cybersecurity features will initially be limited to a vetted group of organizations through the Daybreak Blue program. Broader availability for ChatGPT Plus, Pro, Business, and Enterprise subscribers, as well as API access via Amazon Web Services, is expected in the coming days.

The launch has also drawn scrutiny from security experts regarding a new reasoning technique called 'opaque recurrence' or 'recurrent depth.' This method allows the model to process queries in loops within its latent space, which may make its 'chain of thought' harder for humans to monitor and audit compared to traditional linear reasoning models.

Entities

Amazon Web Services · Amelia Glaese · Astra · Aurora · Buck Shlegeris · CloudSEK · Cursor · Daybreak Blue · Gambit Security · Greg Brockman · Hugging Face · Jakub Pachocki

Claims

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

[● 21 SOURCES] Astra achieved a 100% score on the ExploitBench benchmark for creating exploits for known vulnerabilities. cybersecuritynews.com · www.teknolojia.co.tz · www.blog-nouvelles-technologies.fr · www.fonearena.com · fourweekmba.com · +12 more
[● 19 SOURCES] OpenAI delayed parts of Astra's development to implement additional safety guardrails and monitoring systems. cybersecuritynews.com · www.claimsjournal.com · t3n.de · www.fonearena.com · digitalio.ro · +10 more
[● 32 SOURCES] Astra is the first OpenAI model to reach the 'Critical' cybersecurity capability threshold under the company's Preparedness Framework. cybersecuritynews.com · www.teknolojia.co.tz · www.claimsjournal.com · www.blog-nouvelles-technologies.fr · www.fonearena.com · +18 more
[● 5 SOURCES] In testing, Astra successfully escaped a browser sandbox to execute commands on a host system and achieved root-level access by chaining flaws. cybersecuritynews.com · phocapblockchain.net · cybernoz.com
[● 24 SOURCES] Astra can autonomously identify and exploit security flaws in hardened systems without step-by-step human guidance. cybersecuritynews.com · www.claimsjournal.com · www.fonearena.com · www.deccanchronicle.com · www.teknolojia.co.tz · +14 more
[● 13 SOURCES] During internal testing, Astra discovered and exploited two previously unknown zero-day vulnerabilities. www.teknolojia.co.tz · newscenter.io · cybersecuritynews.com · exame.com · www.hwupgrade.it · +5 more
[● 9 SOURCES] Astra demonstrated superior performance compared to GPT-5.6 in vulnerability identification and exploit development. cybersecuritynews.com · www.fonearena.com · www.newsbytesapp.com · noticiasconcursos.com.br · me.pcmag.com · +3 more
[● 17 SOURCES] OpenAI plans to release Astra soon, initially limiting advanced cyber capabilities to a vetted group through the Daybreak Blue program. www.claimsjournal.com · www.newsbytesapp.com · newscenter.io · t3n.de · fourweekmba.com · +10 more
[● 3 SOURCES] In the Intelligence Index, GPT-6 Astra scored 61 points, matching its predecessor GPT-5.6 Sol. www.trendingtopics.eu · www.gizmodo.jp · forbesliberia.com
[● 8 SOURCES] The model will be available to ChatGPT Plus, Pro, Business, and Enterprise subscribers. bitfinanzas.com · www.elmundo.es · www.adslzone.net · forbesliberia.com · actusmartphone.fr · +3 more
[● 4 SOURCES] OpenAI released GPT-6 Astra, describing it as its most intelligent and aligned model to date. actusmartphone.fr · bitfinanzas.com · forbesliberia.com
[● 2 SOURCES] Anthropic’s Claude Fable 5.1 scored 66 points on the Intelligence Index, outperforming GPT-6 Astra. www.trendingtopics.eu · bitcoinethereumnews.com

Sources

OpenAI pregătește lansarea Astra, un model AI capabil să identifice și să exploateze vulnerabilități informatice [digitalio.ro]

11 days ago

Expertos en seguridad alertan por nueva técnica de razonamiento de OpenAI [espiganoticias.net]

11 days ago

OpenAI Launches GPT-6 Astra After A Curious False Start [forbesliberia.com]

10 days ago

OpenAI’s New Astra AI Can Discover Zero-Day Security Flaws and Build Exploits [cybersecuritynews.com]

12 days ago

OpenAI presentó GPT-6 Astra, un modelo que acerca la IA a la era de los agentes autónomos [www.itnow.connectab2b.com]

10 days ago

Astra của OpenAI đạt ngưỡng nguy hiểm mạng mức nghiêm trọng [phocapblockchain.net]

10 days ago

OpenAI新模型採遞迴推理專家直言「在玩火」憂監控難度升高 [www.technice.com.tw]

11 days ago

OpenAI Yathibitisha Modeli Mpya ya AI Astra Ni Hatari, Inaweza Kuvunja Mifumo. Yaiwekea Ulinzi Mkali. - TeknoKona Teknolojia Tanzania [www.teknolojia.co.tz]

12 days ago

OpenAI Says Upcoming Model Is So Capable It Requires Stronger Guardrails [www.claimsjournal.com]

12 days ago

GPT-6 Astra : OpenAI lance son modèle le plus intelligent, déployé sur ChatGPT et l'API [actusmartphone.fr]

10 days ago

OpenAI's Astra Model Breaks Barriers In Cybersecurity, Raising The Bar For AI Safety [ciente.io]

11 days ago

OpenAI Astra Brings Autonomous Zero-Day Exploitation to AI [cybernoz.com]

11 days ago

OpenAI’den Astra için geri adım! Yeni yapay zeka modelinin çıkışı neden ertelendi? [www.mistikalem.com]

11 days ago

OpenAI Says New Model Meets Its ‘Critical’ Cybersecurity Threshold [retailwit.com]

11 days ago

INTELIGENCIA ARTIFICIAL | OpenAI lanza Astra, su modelo de ChatGPT más potente con autonomía total [www.eldia.es]

10 days ago

OpenAI lancia Astra, addestrato con recurrent depth: nell’API era già comparso il nome gpt-6-astra [ihal.it]

10 days ago

Kritische Cyber-Fähigkeiten: OpenAIs neues KI-Modell Astra kommt an die Leine [t3n.de]

11 days ago

OpenAI anuncia GPT-6 Astra, seu modelo de IA mais inteligente até agora | CNN Brasil [www.cnnbrasil.com.br]

10 days ago

OpenAI : la technique de raisonnement qui inquiète l'IA [www.1001web.fr]

10 days ago

OpenAI lanza Astra, un modelo de IA que maneja el ordenador mejor que un humano y que, admite, cuesta vigilar [www.elmundo.es]

10 days ago

OpenAI ships Astra, hands Daybreak cyber testers first access [bitcoinethereumnews.com]

10 days ago

OpenAI lanza su modelo Astra con advertencias sobre sus avanzadas capacidades cibernéticas [bitfinanzas.com]

10 days ago

Homem x máquina. O risco está cada vez maior [braziljournal.com]

10 days ago

OpenAI trabaja en un sistema de apagado automático para sus modelos de IA [www.diarioestrategia.cl]

10 days ago

OpenAI Warns Its Next AI Model Is So Capable It Needs Stronger Guardrails [newswatchmen.com]

10 days ago

GPT-6 Astra: el nuevo modelo de OpenAI que trabaja por ti [eloutput.com]

10 days ago

OpenAI affirme que son modèle Astra doit être bridé pour sa puissance [www.emarketerz.fr]

11 days ago

Astra hackera meglio di un esperto umano, e fa paura pure a OpenAI [www.hwupgrade.it]

11 days ago

GPT-6 Astra Is Hereand OpenAI Thinks It May Kick Off the AGI Era - IndiaVision India News & Information [www.indiavision.com]

10 days ago

OpenAI dévoile GPT-6 Astra, un nouveau modèle d'IA destiné à des tâches plus complexes. [www.informaticien.be]

10 days ago

GPT-6 Astra: Η OpenAI κυκλοφορεί το «πιο ευφυές μοντέλο τεχνητής νοημοσύνης στον κόσμο» [www.insider.gr]

10 days ago

ChatGPT-6 Astra: OpenAI stärkt Sicherheitsvorkehrungen und Kontrolle [www.it-boltwise.de]

10 days ago

GPT‑6 Astra поставя нови стандарти с рекордни резултати в ARC‑AGI‑3 и ExploitBench - Софтуер и IT Новини [software.kaminata.net]

10 days ago

OpenAI bringt leistungsstärkeres KI-Modell ChatGPT-6 heraus [www.leinetal24.de]

10 days ago

OpenAI Teases Astra, Limits Cybersecurity Tools After Hugging Face Hack [me.pcmag.com]

11 days ago

OpenAI revela novo modelo de IA mais avançada e reforça segurança antes do lançamento [noticiasconcursos.com.br]

11 days ago

OpenAI Presenta Astra, El ChatGPT Más Autónomo Y Potente [www.elperiodico.digital]

10 days ago

INTELIGENCIA ARTIFICIAL | OpenAI lanza Astra, su modelo de ChatGPT más potente con autonomía total [www.diariodemallorca.es]

10 days ago

INTELIGENCIA ARTIFICIAL | OpenAI lanza Astra, su modelo de ChatGPT más potente con autonomía total [www.informacion.es]

10 days ago

INTELIGENCIA ARTIFICIAL | OpenAI lanza Astra, su modelo de ChatGPT más potente con autonomía total [www.farodevigo.es]

10 days ago

10 days ago

OpenAI lança GPT-6 Astra com cibersegurança crítica [www.spacemoney.com.br]

10 days ago

OpenAI declares the “AGI era” with the release of GPT-6 Astra – 4sysops [4sysops.com]

10 days ago

OpenAI lanza un poderoso modelo que augura el inicio la "era de la IA general" [actualidad.rt.com]

10 days ago

KI-Sicherheit: OpenAI-Agenten griffen Hugging Face an – 650 Systeme koordiniert [www.ad-hoc-news.de]

10 days ago

GPT-6 Astra ya es oficial: OpenAI estrena su IA más potente y llegará a ChatGPT en los próximos días [www.adslzone.net]

10 days ago

Open AI priprema napredni model vještačke inteligencije [www.aktuelno.me]

11 days ago

OpenAI’s Astra is its first AI rated ‘Critical’ for hacking, and it found zero-days on its own [astig.ph]

10 days ago

OpenAI a lansat GPT Astra, un model AI despre care susține că "marchează o nouă frontieră în ceea ce privește viteza, precizia și siguranța utilizării computerului". În același timp, compania recunoaște că noul AI este mai predispus să ascundă sau să deghizeze intenționat metodele sale de raționare, ceea ce face mai dificilă evaluarea ulterioară a tehnicilor sale de către oameni. [www.biziday.ro]

10 days ago

OpenAI Astra : la nouvelle IA aux super-pouvoirs de sécurité [www.blog-nouvelles-technologies.fr]

12 days ago

Mythos 5 kompromittiert 15 Systeme: Anthropic reagiert mit neuen Schutzmaßnahmen [borncity.com]

11 days ago

OpenAI cria recurso para desligar sistemas de IA automaticamente [bpmoney.com.br]

11 days ago

OpenAI zet zijn ‘gevaarlijke’ Astra-AI-model toch nu al live [www.bright.nl]

10 days ago

OpenAI launches GPT-6 Astra [www.constellationr.com]

10 days ago

IA: OpenAI trabaja en un sistema de apagado automático para sus modelos contra hackeos » Crónica Viva [www.cronicaviva.com.pe]

10 days ago

Claims of GPT-6 Astra scoring 98. 6% on ARC-AGI-3 don't hold up to scrutiny [cryptobriefing.com]

10 days ago

Astra cruza limiar crítico da guerra cibernética e obriga OpenAI a restringir capacidades avançadas [www.defesanet.com.br]

11 days ago

Lanza OpenAI GPT-6 Astra su modelo de IA más potente [diario.mx]

10 days ago

OpenAI تستعد لإطلاق Astra بقدرات متقدمة لاكتشاف الثغرات السيبرانية [www.elbalad.news]

10 days ago

OpenAI launches Astra, its powerful new AI model [www.emirates247.com]

10 days ago

OpenAI to launch new model with 'stronger safeguards' after hack [www.enca.com]

11 days ago

OpenAI details Astra’s Critical cybersecurity capabilities and safeguards [www.fonearena.com]

12 days ago

OpenAI's Astra Hits 'Critical' Cyber Tier — and Daybreak Blue Turns the Safety Framework Into a Sales Channel - FourWeekMBA [fourweekmba.com]

12 days ago

新AI「GPT-6 Astra」リリース。AGI時代にはキーボードなど叩かない | ギズモード・ジャパン [www.gizmodo.jp]

10 days ago

OpenAI lanza GPT-6 Astra, su IA más potente, pero no te emociones todavía [hipertextual.com]

10 days ago

Nvidia quer ir além dos chips e compra 'biblioteca de IA' por US$ 13 bi [www.uol.com.br]

10 days ago

Servicenow schliesst vier kritische Sicherheitslücken [www.inside-it.ch]

10 days ago

OpenAI's New Reasoning Technique Alarms AI Safety Experts - Slashdot [it.slashdot.org]

10 days ago

INTELIGENCIA ARTIFICIAL | OpenAI lanza Astra, su modelo de ChatGPT más potente con autonomía total [www.elperiodicoextremadura.com]

10 days ago

INTELIGENCIA ARTIFICIAL | OpenAI lanza Astra, su modelo de ChatGPT más potente con autonomía total [www.laopinioncoruna.es]

10 days ago

ChatGPT peut se connecter seul à vos comptes en ligne : un expert alerte sur un risque bien réel [www.letribunaldunet.fr]

10 days ago

OpenAI registra queda global e ChatGPT fica inacessível para milhares de usuários [www.mixvale.com.br]

10 days ago

Astra will be OpenAI's first model with 'critical' cyber abilities [www.newsbytesapp.com]

11 days ago

OpenAI Rates Astra a Critical Cyber Risk. Its Agents Already Took a Cluster. [newscenter.io]

12 days ago

OpenAI adia partes do modelo Astra por risco de cibersegurança [www.ocafezinho.com]

11 days ago

Nova técnica da OpenAI pode tornar seus modelos mais difíceis de monitorar [olhardigital.com.br]

11 days ago

OpenAI adia o lançamento do Astra por ser "demasiado perigoso" para a cibersegurança [pplware.sapo.pt]

11 days ago

OpenAI officially launches GPT-6 Astra, a boundary-breaking new model [sea.mashable.com]

10 days ago

OpenAI Astra Brings Autonomous Zero-Day Exploitation to AI [securityaffairs.com]

11 days ago

Claude Session Hijacking: AI Tool Security for SEO Teams [seovendor.co]

10 days ago

Week of Aug. 31-Sept. 4: What happened, what matters, what's next [www.informationweek.com]

10 days ago

OpenAI insegna ad Astra a ragionare in loop, invisibile a chiunque [www.tomshw.it]

10 days ago

GPT-6 Astra Trails Top Models From Anthropic and Meta in Benchmarks [www.trendingtopics.eu]

10 days ago

Llega GPT-6 Astra, ¿qué puede hacer este potente nuevo modelo de IA y por qué genera preocupación? [www.ultimahora.com]

10 days ago

Researchers fear safety disaster ahead of OpenAI’s Astra release [www.theverge.com]

11 days ago

Künstliche Intelligenz: OpenAI bringt leistungsstärkeres KI-Modell ChatGPT-6 heraus [www.volksfreund.de]

10 days ago

OpenAIの次期モデル「Astra」、サイバー能力が初の“重大”水準に [wired.jp]

10 days ago

Es oficial: GPT-6 Astra ya está aquí e inaugura la nueva era de la Inteligencia Artificial en todo el planeta [www.xataka.com.mx]

10 days ago

OpenAI launches GPT-6 Astra, 'the world’s most intelligent and aligned model' that you can't use just yet [9to5google.com]

10 days ago

Astra: OpenAI pede ajuda às gigantes da segurança para conter nova IA [canaltech.com.br]

11 days ago

OpenAI Says Upcoming Model is So Capable It Requires Stronger Guardrails [www.deccanchronicle.com]

12 days ago

Anthropic Admits Security Failures Behind Claude Hacking Incidents [decrypt.co]

11 days ago

¿ChatGPT diseñará ciberataques? OpenAI lanza Astra con autonomía avanzada [www.elfinanciero.com.mx]

10 days ago

OpenAI limita modelo 'poderoso' após IA atingir nível crítico de risco cibernético [exame.com]

11 days ago

OpenAI limitará el acceso a su inteligencia artificial Astra por su riesgo cibernético "crítico" [www.expansion.com]

11 days ago

Wer KI versteht, hat mehr Angst vor dem Jobverlust [www.faz.net]

10 days ago

OpenAI bringt leistungsstärkeres KI-Modell ChatGPT-6 heraus [www.finanznachrichten.de]

10 days ago

OpenAI begins rollout of new powerful AI model GPT-6 Astra [www.freemalaysiatoday.com]

10 days ago

Katasztrófától tartanak a kutatók az OpenAI terve miatt [hvg.hu]

10 days ago

OpenAI、「GPT-6 Astra」正式発表--数日中にサブスクユーザーへ展開　追加料金不要 - CNET Japan [japan.cnet.com]

10 days ago