Tag: openai
-
OpenAI Agents Collude on Public Wiki to Share Sandbox Bypass and Evasion Techniques
Researchers have discovered a public wiki message board that they claim was used by autonomous AI agents, identifying themselves as OpenAI systems, to exchange answers to tasks, inspect their operating environment, and discuss methods to circumvent sandbox controls. This finding, published on September 4 by Sydney Von Arx, Cormac Slade Byrd, Spencer Kitts, and Thomas…
-
OpenAI Pledges $1bn to Bring its AI Cybersecurity Tools to Essential Services
OpenAI has committed to subsidizing access to Daybreak, helping defenders deploy its AI models in its existing cybersecurity infrastructure First seen on infosecurity-magazine.com Jump to article: www.infosecurity-magazine.com/news/openai-pledges-ai-tools-essential/
-
Cloudflare und OpenAI wollen Schwachstellen schließen, bevor Angreifer sie ausnutzen
Cloudflare verbindet GPT-5.6 Cyber von OpenAI mit Echtzeit-Telemetrie, um Schwachstellen zu priorisieren, Angriffe an der Edge zu blockieren und Patches vorzubereiten. First seen on infopoint-security.de Jump to article: www.infopoint-security.de/cloudflare-und-openai-wollen-schwachstellen-schliessen-bevor-angreifer-sie-ausnutzen/a46341/
-
GPT-6 Astra Scores 100% on ExploitBench as OpenAI Blocks PoC Exploit Requests
OpenAI on Thursday officially unveiled GPT”‘6 Astra, which it described as the “world’s most intelligent and aligned model.”The development comes days after the artificial intelligence (AI) company said the model had reached the “Critical” cybersecurity capability threshold under its Preparedness Framework.”Astra is state-of-the-art on computer use, browsing, software engineering, First seen on thehackernews.com Jump to…
-
OpenAI is putting $1 billion behind Daybreak for defenders working without enterprise budgets
OpenAI committed $1 billion to subsidize access to its Daybreak cyber models, along with training and technical support, for organizations defending water and wastewater … First seen on helpnetsecurity.com Jump to article: www.helpnetsecurity.com/2026/09/04/openai-daybreak-frontline-defenders-access/
-
OpenAI GPT-6 Astra Discovers Zero-Day Flaws and Builds Working Exploits in Cyber Tests
OpenAI has introduced GPT-6 Astra, a groundbreaking model that has reportedly achieved perfect results on ExploitBench and demonstrated improved controls during cybersecurity testing. This announcement positions Astra as a significant advancement in authorized security research, in which models must analyze unfamiliar software vulnerabilities, generate reliable proof-of-concept code, and operate within the parameters of a controlled…
-
OpenAI Launches GPT-6 Astra With Tighter Cyber Safeguards
New Model Can Develop Zero-Day Exploits But Adds More Human Oversight. OpenAI released what it calls its most intelligent and aligned model, GPT-6 Astra, its newest model after it paused some reinforcement training to align its safety and risk processes following agents attacking Hugging Face. First seen on govinfosecurity.com Jump to article: www.govinfosecurity.com/openai-launches-gpt-6-astra-tighter-cyber-safeguards-a-32745
-
Nobody Is Saying Why OpenAI and Anthropic Had Outages Today
ChatGPT, Claude, and Grok all suffered outages at nearly the exact same time for reasons that remain murky. First seen on wired.com Jump to article: www.wired.com/story/nobody-is-saying-why-openai-and-anthropic-had-outages-today/
-
OpenAI confirms ChatGPT is down ahead of ‘Astra’ model launch
ChatGPT and Codex are experiencing a major outage, with users reporting errors across nearly every major ChatGPT feature. First seen on bleepingcomputer.com Jump to article: www.bleepingcomputer.com/news/artificial-intelligence/openai-confirms-chatgpt-is-down-ahead-of-astra-model-launch/
-
OpenAI erweitert Datenresidenz auf ChatGPT Business
Viele mittelständische Unternehmen wollen die Produktivitätsvorteile von KI heute aktiv nutzen, gleichzeitig aber sicherstellen, dass sie ihren internen Anforderungen gerecht werden First seen on infopoint-security.de Jump to article: www.infopoint-security.de/openai-erweitert-datenresidenz-auf-chatgpt-business/a46314/
-
GitSpawn Flaw Enables Arbitrary Code Execution in Claude Code, Codex, Cursor and Grok
A newly disclosed vulnerability class, named GitSpawn, reveals a serious weakness in AI coding agents that automatically execute Git commands to understand a developer’s project better. Researchers at Manifold Security discovered that several tools, including Claude Code, OpenAI Codex, Cursor, Grok Build, Goose, Hermes Agent, and Qwen Code, might inadvertently run commands controlled by an…
-
OpenAI Astra Brings Autonomous Zero-Day Exploitation to AI
OpenAI says Astra can autonomously find zero-days and build exploits, marking its first model to reach the “Critical” cyber risk level. Astra is now officially OpenAI’s highest-risk cybersecurity model. In August, OpenAI said it “couldn’t rule out” that its upcoming model had reached the highest cybersecurity risk level in its Preparedness Framework. In a new…
-
Google, Anthropic, and OpenAI Unveil Cyber AI Models, Safeguards, and Access Programs
Google on Wednesday announced Gemini 3.8 Flash Cyber, which it described as its most capable cybersecurity model, and has made it available to a set of trusted defenders via a new initiative called the Fairwind Program.”The Fairwind Program gives high-priority defenders (like governments, healthcare providers, and telecommunications services) early access to advanced models that help…
-
OpenAI Reveals Astra, Its First AI Model to Reach ‘Critical’ Cybersecurity Risk Threshold
OpenAI announced Tuesday that its upcoming artificial intelligence (AI) model, codenamed Astra, is the company’s first to reach the highest threat level under its internal risk framework, triggering an initial development pause to implement stricter safeguards before its public rollout. Astra attained a >>critical<< designation for cybersecurity capabilities after internal evaluations revealed it could independently..…
-
CrowdStrike Fal.Con 2026: Biggest Statements From Nvidia CEO Huang, Intel CEO Tan And OpenAI President Brockman
At Fal.Con 2026 Tuesday, Nvidia CEO Jensen Huang, Intel CEO Lip-Bu Tan and OpenAI President Greg Brockman joined CrowdStrike co-founder CEO George Kurtz to offer their predictions about emerging cyber risks from agentic AI”, and the need for defenders to rapidly adopt the same technology. First seen on crn.com Jump to article: www.crn.com/news/security/2026/crowdstrike-fal-con-2026-biggest-statements-from-nvidia-ceo-huang-intel-ceo-tan-and-openai-president-brockman
-
KI agierte eigenständig – OpenAI stärkt nach KI-Hacks Schutzvorkehrungen bei Tests
First seen on security-insider.de Jump to article: www.security-insider.de/openai-verschaerft-ki-sicherheit-nach-test-hack-a-bfdcd02eb24ed2593ae4a63d0f725c52/
-
OpenAI Is About to Release Its First AI Model With ‘Critical’ Cyber Abilities
The company will give select partners early access to its Astra AI model”, so they have time to shore up their defenses. First seen on wired.com Jump to article: www.wired.com/story/openai-astra-first-ai-model-with-critical-cyber-abilities/
-
Critical Langflow flaw exploited to steal OpenAI and AWS keys
Tags: ai, credentials, exploit, flaw, framework, open-source, openai, remote-code-execution, threat, vulnerabilityThreat actors are exploiting an unauthenticated remote code execution vulnerability (CVE-2026-0768) in Langflow, an open-source framework for building AI applications, to steal credentials, tokens, and keys. First seen on bleepingcomputer.com Jump to article: www.bleepingcomputer.com/news/security/critical-langflow-flaw-exploited-to-steal-openai-and-aws-keys/
-
What Does a SOC Look Like When AI Is Part of the Architecture? A Closed-Door Working Session on September 15
Alert volumes are climbing. And in July, the theoretical version of the AI threat stopped being theoretical. An autonomous agent framework driven by OpenAI models ran an end-to-end intrusion against Hugging Face, executing roughly 17,600 recorded actions across a four… First seen on securityboulevard.com Jump to article: securityboulevard.com/2026/09/what-does-a-soc-look-like-when-ai-is-part-of-the-architecture-a-closed-door-working-session-on-september-15/
-
Hackers Exploit Langflow RCE Flaw to Harvest OpenAI and AWS Credentials
Tags: ai, credentials, cve, cyber, exploit, flaw, hacker, openai, rce, remote-code-execution, threat, vulnerabilityThreat actors are actively exploiting a critical remote code execution vulnerability in Langflow, a low-code platform used for building AI-powered applications and automating workflows. This vulnerability, tracked as CVE-2026-0768, affects the code validator in Langflow’s custom component editor. According to Caitlin Condon, VP of Security Research at VulnCheck, the flaw allows unauthenticated remote code execution…
-
Fake OpenAI, Anthropic and DeepSeek Crawlers Target .env Files and Cloud Credentials
Threat actors are impersonating AI web crawlers from organizations such as OpenAI, Anthropic, DeepSeek, Google, Perplexity, and Amazon to scan internet-facing servers for exposed secrets, according to a GreyNoise research report published on August 28, 2026. This activity involves automated scanners that use forged crawler user-agent strings to request sensitive files, including .env configurations, AWS…
-
OpenAI Cuts Off Cursor Models After SpaceX Acquisition
OpenAI Cites Distrust of Musk as Cursor Faces Growing Pressure From Rival Coding Tools. After Cursor’s sale to Elon Musk’s SpaceX, OpenAI announced it will gradually cut off access to its models on the vibe-coding platform, citing distrust of the company’s new owner. First seen on govinfosecurity.com Jump to article: www.govinfosecurity.com/openai-cuts-off-cursor-models-after-spacex-acquisition-a-32700
-
AI Model Rules Are Not Security Controls
OpenAI’s Hugging Face attack postmortem shows agents don’t care about rules, they need strong controls. First seen on darkreading.com Jump to article: www.darkreading.com/cyber-risk/model-knowing-rules-is-not-security-control
-
OpenAI confirms ChatGPT outage as users report errors
ChatGPT Work is experiencing a partial outage, and users across multiple subscription plans may be unable to start or continue tasks. First seen on bleepingcomputer.com Jump to article: www.bleepingcomputer.com/news/artificial-intelligence/openai-confirms-chatgpt-outage-as-users-report-errors/
-
Threat actors are posing as AI crawlers to hunt for exposed credentials
Attackers are disguising automated scanning as traffic from AI crawlers operated by OpenAI, Anthropic, Google, Perplexity and other companies while searching websites for … First seen on helpnetsecurity.com Jump to article: www.helpnetsecurity.com/2026/08/31/ai-crawlers-scan-exposed-credentials/
-
OpenAI and 100+ Firms Warn AI Cyberattacks Could Surge Within Months
More than 100 organizations warn AI could accelerate cyberattacks faster than defenders can respond, putting critical infrastructure at greater risk. First seen on esecurityplanet.com Jump to article: www.esecurityplanet.com/threats/news-openai-ai-cyber-defense-critical-infrastructure/
-
OpenAI Warns Astra AI Model May Develop Zero-Day Exploits and Launch Autonomous Cyberattacks
Tags: ai, cyber, cyberattack, cybersecurity, exploit, intelligence, openai, update, vulnerability, zero-dayOpenAI has issued a warning regarding Astra, an upcoming artificial intelligence model, which may be close to a threshold of cybersecurity capabilities that would allow it to independently discover zero-day vulnerabilities and conduct complex cyberattacks against highly secured systems. In a security update dated August 7, 2026, the company noted that early testing and assessments…
-
28,000 Exposed Git Repositories Found Leaking Credentials
Researchers found 28,000 exposed Git repositories containing active AWS, Stripe, OpenAI and GitHub credentials. Learn the risks and defenses. First seen on esecurityplanet.com Jump to article: www.esecurityplanet.com/threats/news-28000-exposed-git-repositories-leak-credentials/
-
AI Doesn’t Mean the End of Mathematics”, at Least Not Yet
This essay was written with Kasra Rafi, and originally appeared in The Guardian. Earlier this month, about 40 top mathematicians gathered at OpenAI’s offices to discuss the future of their profession. The meeting was off-the-record, but if recent articles by mathematicians are any guide, it was mostly pretty glum. People fear for their jobs, their…
-
700 OpenAI Agents Coordinate Attack on Hugging Face and Gain Remote Code Execution
OpenAI’s ExploitGym evaluation environment reportedly became the site of a large-scale, unsanctioned multi-agent campaign after hundreds of models found ways to communicate across supposedly isolated sandboxes. An investigation published by METR describes how the activity, which began on July 8, involved agents operating across several models, including GPT-5.6 Sol and an internally persistent model identified…