Tag: openai
-
OpenAI Warns Organizations to Automate Cybersecurity as AI-Powered Attacks Accelerate
OpenAI has issued a warning that organizations need to quickly automate core cybersecurity functions as increasingly advanced AI systems make it easier and cheaper to identify, exploit, and chain security vulnerabilities. In a recent security article titled “The Defender’s Window,” OpenAI President Greg Brockman explained that the OpenAI-Hugging Face incident showcased how highly capable attackers…
-
Adam Shostack Talks Hugging Face Breach & PHANTOM-B
The security expert talks with the Dark Reading News Desk about why he blown away by OpenAI’s revelations regarding the Hugging Face attack, and also discussed his new threat model for LLMs. First seen on darkreading.com Jump to article: www.darkreading.com/vulnerabilities-threats/adam-shostack-talks-hugging-face-phantom-b
-
OpenAI tightens defenses after AI agents breach research environment
Following the OpenAI-Hugging Face incident, in which an agentic collective autonomously penetrated OpenAI’s research infrastructure and another company’s production … First seen on helpnetsecurity.com Jump to article: www.helpnetsecurity.com/2026/08/18/openai-strengthening-security-measures/
-
Hackers Turn Claude Code and Codex Into AI-Powered Tools for Credential Theft and Cloud Attacks
Threat actors are increasingly using coding assistants as operational tools. Detailed research from Gambit Security highlights three campaigns where Claude Code, OpenAI Codex, and large language models facilitated activities ranging from ransomware preparation to the harvesting of secrets on a large scale and exploiting cloud accounts. These cases demonstrate how AI can speed up attackers’…
-
OpenAI President Urges Enterprises to Deploy AI Agents
Greg Brockman Says Defenders Must Automate Security Before Attackers Gain Ground. OpenAI president Greg Brockman believes every enterprise should bring agents to its security teams as urgently as possible to combat sophisticated attacks, even after its own agents were responsible for an attack against model repository Hugging Face. First seen on govinfosecurity.com Jump to article:…
-
OpenAI President Urges Enterprises to Deploy AI Agents
Greg Brockman Says Defenders Must Automate Security Before Attackers Gain Ground. OpenAI president Greg Brockman believes every enterprise should bring agents to its security teams as urgently as possible to combat sophisticated attacks, even after its own agents were responsible for an attack against model repository Hugging Face. First seen on govinfosecurity.com Jump to article:…
-
OpenAI Workload Identity Federation: Aembit Brings Secretless Access to the OpenAI API
3 min readAembit already covers a lot of ground when it comes to securing AI workload access. For OpenAI’s ChatGPT, workloads can authenticate to the ChatGPT API using static API key injection, with the Aembit proxy handling direct access transparently. Today, we’re extending that coverage with the introduction of the OpenAI Workload Identity Federation Credential…
-
Adam Shostack Talks Hugging Face & PHANTOM-B
World-class threat modeler Adam Shostack shared he was blown away by OpenAI’s revelations about the Hugging Face attack, and explains why his new threat model for LLMs is both lightweight yet still usable. First seen on darkreading.com Jump to article: www.darkreading.com/vulnerabilities-threats/adam-shostack-talks-hugging-face-phantom-b
-
OpenAI, Anthropic, and Meta AI Breaches Shared the Same Testing Vendor
OpenAI, Anthropic, and Meta AI incidents reportedly shared one testing vendor, exposing third-party and containment risks in AI security evaluations. First seen on esecurityplanet.com Jump to article: www.esecurityplanet.com/cloud-security/news-openai-anthropic-meta-ai-incidents-irregular/
-
Verschlüsselte KI-Denkprotokolle geknackt Schwachstelle bei OpenAI, Anthropic und Google
Forscher haben eine gravierende Schwachstelle bei der Absicherung sogenannter Reasoning-Logs entdeckt. Verschlüsselte Denkprotokolle moderner KI-Modelle lassen sich demnach unter bestimmten Bedingungen über ein schwächeres Modell desselben Anbieters entschlüsseln. Besonders brisant: In öffentlich zugänglichen Datensätzen fanden die Forscher bereits personenbezogene Daten, Zugangsdaten, API-Schlüssel und Passwörter. Forscher von MATS Research, dem Max-Planck-Institut für intelligente Systeme, dem ELLIS…
-
Drei KI-Vorfälle in vierzehn Tagen Von der Evaluierung zum Ernstfall
Innerhalb von vierzehn Tagen haben OpenAI, Anthropic und das britische AI Security Institute (AISI) jeweils offengelegt, dass KI-Agenten im Rahmen interner Sicherheitsprüfungen den vorgesehenen Testrahmen verlassen und auf reale Systeme sowie reale Personen eingewirkt haben. Weniger bemerkenswert als die Einzelfälle ist dabei die Geschwindigkeit, mit der sich die Fähigkeiten dieser Agenten entwickeln und der […]…
-
OpenAI Launches GPT-5.6-Cyber with Reduced Safeguards for Exploit Development
OpenAI on Monday unveiled a new cybersecurity-focused model called GPT”‘5.6″‘Cyber that it said is focused on vulnerability research, penetration testing, and incident response.”Built on GPT”‘5.6 Sol, it is trained to improve capabilities on several specialized cybersecurity tasks (e.g., finding zero-day vulnerabilities and developing exploit chains) and to reduce refusals for certain higher-risk First seen on…
-
OpenAI Launches Two-Tier Security Access Program Alongside GPT 5.6 Cyber
Daybreak Blue removes some OpenAI-made guardrails while Daybreak Red grants the use of cyber-focused frontier AI models First seen on infosecurity-magazine.com Jump to article: www.infosecurity-magazine.com/news/openai-daybreak-blue-red-gpt-cyber/
-
OpenAI Pauses Some Development of Astra Model on Security Concerns
Tags: openaiOpenAI is tightening restrictions on testing of its upcoming Astra model due to security concerns First seen on infosecurity-magazine.com Jump to article: www.infosecurity-magazine.com/news/openai-pauses-development-astra/
-
OpenAI, Anthropic und AISI: Was die drei KI-Vorfälle über Agentensicherheit aussagen
Drei KI-Sicherheitsvorfälle bei OpenAI, Anthropic und AISI zeigen, warum Unternehmen Kontrolle, Monitoring und Governance für KI-Agenten benötigen. First seen on infopoint-security.de Jump to article: www.infopoint-security.de/openai-anthropic-und-aisi-was-die-drei-ki-vorfaelle-ueber-agentensicherheit-aussagen/a46084/
-
GPT-5.6-Cyber refuses security researchers’ requests far less often
GPT-5.6-Cyber is a new OpenAI model built on GPT-5.6 Sol, trained to find zero-day vulnerabilities and build exploit chains, with fewer refusals on higher-risk, dual-use work. … First seen on helpnetsecurity.com Jump to article: www.helpnetsecurity.com/2026/08/11/openai-gpt-5-6-cyber-model/
-
OpenAI Launches GPT-5.6-Cyber to Find Zero-Day Vulnerabilities and Develop Exploit Chains
OpenAI has expanded its Daybreak cybersecurity program with the introduction of GPT-5.6-Cyber, a purpose-trained model specifically designed for authorized vulnerability research, exploit validation, and advanced security testing. Built on the foundation of GPT-5.6 Sol, this new model serves as a controlled-access tool for trusted defenders as AI-assisted offensive capabilities continue to evolve. GPT-5.6-Cyber to Find…
-
Your security vendor gets the frontier cyber model, you get the findings
Selected red team specialists can now use OpenAI’s cyber models to find and exploit weaknesses in client applications and infrastructure. Those clients never get the … First seen on helpnetsecurity.com Jump to article: www.helpnetsecurity.com/2026/08/11/openai-daybreak-cyber-models/
-
AI Moves From Cheating in Theory to Hacking the Real World
Why Zero Trust for AI and Enforced Hard Constraints Beat Easily Ignored Rules Among the many lessons learned from the OpenAI sandbox escape/Hugging Face hack and the more recent Claude accidental escape is that artificial intelligence cheats: It breaks rules then denies it. These aren’t anomalies. They’re fundamental systemic failures. First seen on govinfosecurity.com Jump…
-
OpenAI says Daybreak will expand to offer specialized cyber services
The company rolled out “Red” and “Blue” programs for defenders, introduced a new model and announced partnerships with 16 major cybersecurity vendors. First seen on cyberscoop.com Jump to article: cyberscoop.com/openai-daybreak-expansion-specialized-cyber-services/
-
OpenAI releases ChatGPT 5.6 Cyber, but it’s only for approved users
OpenAI has developed a new model called “GPT 5.6 Cyber,” designed for vulnerability research, penetration testing, incident response, and remediation. First seen on bleepingcomputer.com Jump to article: www.bleepingcomputer.com/news/security/openai-releases-chatgpt-56-cyber-but-its-only-for-approved-users/
-
Huntress CEO: The Autonomous Adversary Is Here And ‘We’ve All Been Drafted’
The autonomous compromise recently disclosed by OpenAI shows clearly that the AI-powered threats that security experts have been warning about are moving from theory into reality, Huntress CEO Kyle Hanslovan said during the latest episode of CRN’s Security or Else! First seen on crn.com Jump to article: www.crn.com/news/security/2026/huntress-ceo-the-autonomous-adversary-is-here-and-we-ve-all-been-drafted
-
OpenAI Pauses Astra Model Over Critical Cybersecurity Risk Concerns
OpenAI paused work involving Astra after tests showed cybersecurity abilities that could approach its Critical risk threshold under the company’s framework. OpenAI disclosed that internal evaluations of Astra, one of its upcoming models, have found cybersecurity capabilities significant enough that the company >>cannot rule out<< reaching the Critical threshold under its own Preparedness Framework. In…
-
OpenAI locks down Astra over potential critical cyber capabilities
OpenAI’s internal evaluation of its upcoming model, Astra, found significant advances in agentic coding and cybersecurity, leading the company to conclude that it cannot … First seen on helpnetsecurity.com Jump to article: www.helpnetsecurity.com/2026/08/10/openai-astra-critical-cyber-capabilities/
-
OpenAI’s Next AI Model Astra Shows Cyber Performance Strong Enough to Trigger Pause
OpenAI has announced that it’s pausing some “internal activities” involving its upcoming artificial intelligence (AI) model Astra after an internal evaluation found it had made significant advancements in agentic coding and cybersecurity.In response to the discovery, the AI upstart said it’s implementing security controls for higher-capability models and associated activities, such as isolated First seen…
-
OpenAI Pauses Development on Powerful Astra Model Over Autonomous Cyberattack Risks
OpenAI has halted select internal development on its unreleased flagship model, Astra, after safety evaluations revealed the system had achieved unprecedented autonomous cyberattack capabilities. The move comes as the artificial intelligence (AI) industry grapples with a wave of containment failures, where increasingly autonomous models have repeatedly breached sandbox environments and accessed live targets on the..…
-
Astra: OpenAI will neues KI-Modell vorerst nicht veröffentlichen
Wie gefährlich darf eine KI werden, bevor sie zur Waffe wird? OpenAI kann diese Frage bei seinem Modell Astra derzeit nicht mehr sicher beantworten. First seen on golem.de Jump to article: www.golem.de/news/astra-openai-will-neues-ki-modell-vorerst-nicht-veroeffentlichen-2608-211738.html
-
Nach OpenAI und Anthropic: Auch chinesisches KI-Modell aus Testumgebung ausgebrochen
First seen on t3n.de Jump to article: t3n.de/news/chinesisches-ki-modell-kimi-k3-aus-testumgebung-ausgebrochen-1757100/
-
OpenAI disrupts major ChatGPT-driven scam campaign in Cambodia
First seen on scworld.com Jump to article: www.scworld.com/brief/openai-disrupts-major-scam-campaign-in-cambodia-using-chatgpt
-
Déjà Vu? Meta’s AI Escapes Testing Lab in Hacking Joyride
In the span of three weeks, OpenAI, Anthropic, and Meta have all disclosed AI agent sandbox escape events affecting real organizations. First seen on darkreading.com Jump to article: www.darkreading.com/cyberattacks-data-breaches/meta-ai-escapes-lab-hacking-joyride