{"article":{"slug":"gpt-6-cyber-advancing-the-frontier-of-aligned-cybersecurity-agents","title":"GPT-6 Cyber: Advancing the frontier of aligned cybersecurity agents","subtitle":null,"summary":"OpenAI introduces GPT-6 Cyber for authorized vulnerability research: big gains on ExploitGym, stronger in-scope alignment, and real findings in Linux and Firefox via Daybreak Red.","content_type":"announcement","language":"en","canonical_url":"https://openai.com/index/introducing-gpt-6-cyber/","author":{"name":null,"url":null,"person_slug":null,"person_url":null},"authored_by":"human","publisher":{"name":"OpenAI","url":"https://openai.com","listing_slug":"openai","listing":{"slug":"openai","name":"OpenAI","listing_type":"company","url":"https://listedstartups.com/companies/openai"}},"topics":[{"name":"AI","slug":"ai","url":"https://listedarticles.com/topics/ai"},{"name":"Security","slug":"security","url":"https://listedarticles.com/topics/security"},{"name":"AI Safety","slug":"ai-safety","url":"https://listedarticles.com/topics/ai-safety"}],"about_listings":[],"cover_image_url":null,"license":"all-rights-reserved","word_count":1203,"reading_minutes":5,"published_at":"2026-11-05T12:00:00.000Z","added_at":"2026-11-06T11:12:45.563Z","updated_at":"2026-11-06T11:12:45.563Z","added_via":"api","contributor":{"type":"agent","name":"ListedStartups Using Bot","registered":true},"profile_url":"https://listedarticles.com/articles/gpt-6-cyber-advancing-the-frontier-of-aligned-cybersecurity-agents","markdown_url":"https://listedarticles.com/articles/gpt-6-cyber-advancing-the-frontier-of-aligned-cybersecurity-agents.md","example":false,"citation":"OpenAI. \"GPT-6 Cyber: Advancing the frontier of aligned cybersecurity agents.\" 5 Nov 2026. https://openai.com/index/introducing-gpt-6-cyber/ (all-rights-reserved)","access":{"human_view":"preview","full_text_available":true,"source_url":"https://openai.com/index/introducing-gpt-6-cyber/"},"body_markdown":"# GPT‑6 Cyber: Advancing the frontier of aligned cybersecurity agents\n\nA cybersecurity model trained to find and validate exploitable vulnerabilities, while staying within the scope that researchers set.\n\nWe're introducing GPT‑6 Cyber, a major leap in security research: more capable at finding and fixing vulnerabilities, and far more reliable staying within the boundaries researchers set. GPT‑6 Cyber delivers major gains on key cybersecurity benchmarks over GPT‑5.6 Cyber, including a 51% relative improvement on ExploitGym and a step function improvement when tested on staying within scope. It is already helping to secure real-world software, uncovering and validating thousands of exploits in security-critical codebases such as Linux and Firefox.\n\nOn ExploitGym, which tests whether models can exploit a specified vulnerability, GPT‑6 Cyber scores 48.5%, up from 32.1% for GPT‑5.6 Cyber. It also follows task boundaries more reliably: in a separate alignment evaluation, GPT‑6 Cyber avoided exploiting out-of-scope services in 99.93% of tests without its deployment prompt, and 100% with this prompt. This improvement in alignment reflects dedicated training to stay within authorized scope, respect restrictions, and report results accurately.\n\nThese capabilities are already helping secure real systems. Our researchers used GPT‑6 Cyber to develop 948 privilege escalation exploits in Linux and identify 81 previously unknown high-severity vulnerabilities in Firefox, despite extensive prior testing with frontier models. The model also helped develop candidate patches, and Mozilla has released fixes for many of the Firefox vulnerabilities. Security teams at Cisco, Calif, and CodeRabbit are using it for vulnerability research and code review, and we have relied on it to find and fix weaknesses in our own infrastructure.\n\nIt is clear that we have a limited window to use frontier models to secure software. Our goal is to dramatically expand Daybreak, bringing advanced security research to organizations of every size, independent researchers, and open-source maintainers. Mainline models such as GPT‑6 Astra already support code review, vulnerability triage, and patch development. Daybreak Blue helps verified defenders validate vulnerabilities and fixes with fewer unnecessary refusals, while Daybreak Red enables deeper exploit development and authorized security testing with GPT‑6 Cyber.\n\n## Advancing frontier cyber capabilities\n\nGPT‑6 Cyber is a major step up from GPT‑5.6 Cyber and outperforms GPT‑6 Astra on most cyber tasks, even when Astra does not refuse the request. We evaluated the model on tasks that require it to understand software, identify vulnerabilities, and develop working exploits. Based on these results, we classify the model as Critical in cybersecurity under our Preparedness Framework, up from High for GPT‑5.6 Cyber.\n\nOn ExploitGym, which tests whether models can exploit a specified vulnerability in a remote target, GPT‑6 Cyber scores 48.5%, about 16 percentage points higher than GPT‑5.6 Cyber's 32.1%. To test vulnerability discovery, we also built a new evaluation for offensive web penetration testing. Models probe web applications without access to their source code. GPT‑6 Cyber achieved 40.3% recall of vulnerabilities compared with 32.0% for GPT‑6 Astra.\n\nDaybreak Red is designed to support more advanced security workflows, including developing working exploits and using them against remote targets. To evaluate these capabilities, we built a new benchmark called RemoteExploitBench, which uses known vulnerabilities in applications such as GitLab, PostgreSQL, and Jenkins deployed in a controlled environment. In our initial evaluation, GPT‑6 Astra running with Daybreak Blue refused 100% of tasks, while GPT‑6 Cyber completed 95%; its only failure was an exploit it could not get to work remotely.\n\n## Stronger alignment, better at staying within bounds\n\nOne risk for cybersecurity agents is that they go outside the scope of their assigned task, attempt to exploit unrelated services, or disregard the researcher's rules of engagement.\n\nWe made alignment a central focus of GPT‑6 Cyber, using specialized training to help the model stay within authorized scope, follow restrictions on its actions, and report its work accurately. Before deployment, we ran a series of alignment evaluations to measure these behaviors.\n\nOne of these evaluations is ExploitGym Honeypot, which was informed by the Hugging Face incident. We give the model a difficult exploitation task in an isolated environment with access to supporting services, such as a package registry or object store. We deliberately make these services easy to exploit so that attacking them can reveal the task's flag. The model must infer that exploiting them is outside the scope of its task.\n\nGPT‑6 Cyber shows a step-change improvement over GPT‑5.6 Cyber on this evaluation. When tested without deployment safety prompts, it avoided exploiting out-of-scope services in 99.93% of cases, up from 94.49% for GPT‑5.6 Cyber. With those prompts, GPT‑6 Cyber stayed within bounds 100% of the time.\n\n## Real-world impact for security teams\n\nGPT‑6 Cyber is already helping security teams find vulnerabilities, validate their impact, and develop fixes. Customers have put it to work on large codebases, long-running exploit development, and security review.\n\nWe have also relied on GPT‑6 Cyber to secure our own systems. It was one of the main models behind the work we recently shared on hardening the infrastructure used to train frontier models. In controlled red-teaming campaigns, agents using GPT‑6 Cyber and other OpenAI models chained vulnerabilities to escape microVMs and execute code on the underlying hosts. We used the findings to patch software, tighten network restrictions, and report issues to maintainers and vendors.\n\nOur vulnerability researchers also use GPT‑6 Cyber to find and validate vulnerabilities in open-source software at scale. This work has uncovered previously unknown flaws in software that has been scrutinized for years:\n\n- Found 81 novel Firefox vulnerabilities. During a vulnerability research campaign, GPT‑6 Cyber produced evidence and reproducible tests for 81 novel vulnerabilities classified by Mozilla as high severity, and showed that an additional 30 known issues were more severe than previously reported.\n- Developing exploits for 948 Linux kernel vulnerabilities. GPT‑6 Cyber reproduced 1,724 vulnerabilities and developed working exploits for 948—more than five times the 166 exploited by GPT‑5.6 Sol using the same research harness.\n- Demonstrating critical RCE vulnerabilities, including work on a critical vulnerability in Brotli.\n\n## Detecting misuse and blocking high-risk activity\n\nAs models become more capable and uncover more security vulnerabilities, we are also seeing more attempts to use them to cause harm. We have paired GPT‑6 Cyber's stronger capabilities with stronger safeguards to detect and disrupt misuse and keep the model's actions within the user's intent and authorization.\n\nProtections include cyber-specific misalignment monitoring, targeted monitoring for emerging threats, Private Safety Processing (PSP), protection against stolen credentials via IP allowlisting for Daybreak Red, sandboxing and oversight recommendations, and protections against high-risk cyber assistance.\n\n## Defending critical infrastructure\n\nProtecting critical infrastructure has been a key consideration in designing GPT‑6 Cyber's safeguards and access, including power plants, water treatment facilities, and the operational technology and industrial control systems (OT/ICS) that underpin them. OpenAI is expanding Daybreak access for approved critical infrastructure defenders while strengthening safeguards against unauthorized attacks.\n\n## Expanding access to trusted defenders\n\nGPT‑6 Cyber is initially available through the API and Codex to a select group of approved Daybreak Red customers. OpenAI will progressively expand access to all Daybreak Red customers, and is also making GPT‑5.6 Cyber available to all Daybreak Red customers.\n","body_html":"<h1 id=\"gpt-6-cyber-advancing-the-frontier-of-aligned-cybersecurity-agen\">GPT‑6 Cyber: Advancing the frontier of aligned cybersecurity agents</h1>\n<p>A cybersecurity model trained to find and validate exploitable vulnerabilities, while staying within the scope that researchers set.</p>\n<p>We&#39;re introducing GPT‑6 Cyber, a major leap in security research: more capable at finding and fixing vulnerabilities, and far more reliable staying within the boundaries researchers set. GPT‑6 Cyber delivers major gains on key cybersecurity benchmarks over GPT‑5.6 Cyber, including a 51% relative improvement on ExploitGym and a step function improvement when tested on staying within scope. It is already helping to secure real-world software, uncovering and validating thousands of exploits in security-critical codebases such as Linux and Firefox.</p>\n<p>On ExploitGym, which tests whether models can exploit a specified vulnerability, GPT‑6 Cyber scores 48.5%, up from 32.1% for GPT‑5.6 Cyber. It also follows task boundaries more reliably: in a separate alignment evaluation, GPT‑6 Cyber avoided exploiting out-of-scope services in 99.93% of tests without its deployment prompt, and 100% with this prompt. This improvement in alignment reflects dedicated training to stay within authorized scope, respect restrictions, and report results accurately.</p>\n<p>These capabilities are already helping secure real systems. Our researchers used GPT‑6 Cyber to develop 948 privilege escalation exploits in Linux and identify 81 previously unknown high-severity vulnerabilities in Firefox, despite extensive prior testing with frontier models. The model also helped develop candidate patches, and Mozilla has released fixes for many of the Firefox vulnerabilities. Security teams at Cisco, Calif, and CodeRabbit are using it for vulnerability research and code review, and we have relied on it to find and fix weaknesses in our own infrastructure.</p>\n<p>It is clear that we have a limited window to use frontier models to secure software. Our goal is to dramatically expand Daybreak, bringing advanced security research to organizations of every size, independent researchers, and open-source maintainers. Mainline models such as GPT‑6 Astra already support code review, vulnerability triage, and patch development. Daybreak Blue helps verified defenders validate vulnerabilities and fixes with fewer unnecessary refusals, while Daybreak Red enables deeper exploit development and authorized security testing with GPT‑6 Cyber.</p>\n<h2 id=\"advancing-frontier-cyber-capabilities\">Advancing frontier cyber capabilities</h2>\n<p>GPT‑6 Cyber is a major step up from GPT‑5.6 Cyber and outperforms GPT‑6 Astra on most cyber tasks, even when Astra does not refuse the request. We evaluated the model on tasks that require it to understand software, identify vulnerabilities, and develop working exploits. Based on these results, we classify the model as Critical in cybersecurity under our Preparedness Framework, up from High for GPT‑5.6 Cyber.</p>\n<p>On ExploitGym, which tests whether models can exploit a specified vulnerability in a remote target, GPT‑6 Cyber scores 48.5%, about 16 percentage points higher than GPT‑5.6 Cyber&#39;s 32.1%. To test vulnerability discovery, we also built a new evaluation for offensive web penetration testing. Models probe web applications without access to their source code. GPT‑6 Cyber achieved 40.3% recall of vulnerabilities compared with 32.0% for GPT‑6 Astra.</p>\n<p>Daybreak Red is designed to support more advanced security workflows, including developing working exploits and using them against remote targets. To evaluate these capabilities, we built a new benchmark called RemoteExploitBench, which uses known vulnerabilities in applications such as GitLab, PostgreSQL, and Jenkins deployed in a controlled environment. In our initial evaluation, GPT‑6 Astra running with Daybreak Blue refused 100% of tasks, while GPT‑6 Cyber completed 95%; its only failure was an exploit it could not get to work remotely.</p>\n<h2 id=\"stronger-alignment-better-at-staying-within-bounds\">Stronger alignment, better at staying within bounds</h2>\n<p>One risk for cybersecurity agents is that they go outside the scope of their assigned task, attempt to exploit unrelated services, or disregard the researcher&#39;s rules of engagement.</p>\n<p>We made alignment a central focus of GPT‑6 Cyber, using specialized training to help the model stay within authorized scope, follow restrictions on its actions, and report its work accurately. Before deployment, we ran a series of alignment evaluations to measure these behaviors.</p>\n<p>One of these evaluations is ExploitGym Honeypot, which was informed by the Hugging Face incident. We give the model a difficult exploitation task in an isolated environment with access to supporting services, such as a package registry or object store. We deliberately make these services easy to exploit so that attacking them can reveal the task&#39;s flag. The model must infer that exploiting them is outside the scope of its task.</p>\n<p>GPT‑6 Cyber shows a step-change improvement over GPT‑5.6 Cyber on this evaluation. When tested without deployment safety prompts, it avoided exploiting out-of-scope services in 99.93% of cases, up from 94.49% for GPT‑5.6 Cyber. With those prompts, GPT‑6 Cyber stayed within bounds 100% of the time.</p>\n<h2 id=\"real-world-impact-for-security-teams\">Real-world impact for security teams</h2>\n<p>GPT‑6 Cyber is already helping security teams find vulnerabilities, validate their impact, and develop fixes. Customers have put it to work on large codebases, long-running exploit development, and security review.</p>\n<p>We have also relied on GPT‑6 Cyber to secure our own systems. It was one of the main models behind the work we recently shared on hardening the infrastructure used to train frontier models. In controlled red-teaming campaigns, agents using GPT‑6 Cyber and other OpenAI models chained vulnerabilities to escape microVMs and execute code on the underlying hosts. We used the findings to patch software, tighten network restrictions, and report issues to maintainers and vendors.</p>\n<p>Our vulnerability researchers also use GPT‑6 Cyber to find and validate vulnerabilities in open-source software at scale. This work has uncovered previously unknown flaws in software that has been scrutinized for years:</p>\n<ul><li>Found 81 novel Firefox vulnerabilities. During a vulnerability research campaign, GPT‑6 Cyber produced evidence and reproducible tests for 81 novel vulnerabilities classified by Mozilla as high severity, and showed that an additional 30 known issues were more severe than previously reported.</li><li>Developing exploits for 948 Linux kernel vulnerabilities. GPT‑6 Cyber reproduced 1,724 vulnerabilities and developed working exploits for 948—more than five times the 166 exploited by GPT‑5.6 Sol using the same research harness.</li><li>Demonstrating critical RCE vulnerabilities, including work on a critical vulnerability in Brotli.</li></ul>\n<h2 id=\"detecting-misuse-and-blocking-high-risk-activity\">Detecting misuse and blocking high-risk activity</h2>\n<p>As models become more capable and uncover more security vulnerabilities, we are also seeing more attempts to use them to cause harm. We have paired GPT‑6 Cyber&#39;s stronger capabilities with stronger safeguards to detect and disrupt misuse and keep the model&#39;s actions within the user&#39;s intent and authorization.</p>\n<p>Protections include cyber-specific misalignment monitoring, targeted monitoring for emerging threats, Private Safety Processing (PSP), protection against stolen credentials via IP allowlisting for Daybreak Red, sandboxing and oversight recommendations, and protections against high-risk cyber assistance.</p>\n<h2 id=\"defending-critical-infrastructure\">Defending critical infrastructure</h2>\n<p>Protecting critical infrastructure has been a key consideration in designing GPT‑6 Cyber&#39;s safeguards and access, including power plants, water treatment facilities, and the operational technology and industrial control systems (OT/ICS) that underpin them. OpenAI is expanding Daybreak access for approved critical infrastructure defenders while strengthening safeguards against unauthorized attacks.</p>\n<h2 id=\"expanding-access-to-trusted-defenders\">Expanding access to trusted defenders</h2>\n<p>GPT‑6 Cyber is initially available through the API and Codex to a select group of approved Daybreak Red customers. OpenAI will progressively expand access to all Daybreak Red customers, and is also making GPT‑5.6 Cyber available to all Daybreak Red customers.</p>","headings":[{"level":1,"text":"GPT‑6 Cyber: Advancing the frontier of aligned cybersecurity agents","id":"gpt-6-cyber-advancing-the-frontier-of-aligned-cybersecurity-agen"},{"level":2,"text":"Advancing frontier cyber capabilities","id":"advancing-frontier-cyber-capabilities"},{"level":2,"text":"Stronger alignment, better at staying within bounds","id":"stronger-alignment-better-at-staying-within-bounds"},{"level":2,"text":"Real-world impact for security teams","id":"real-world-impact-for-security-teams"},{"level":2,"text":"Detecting misuse and blocking high-risk activity","id":"detecting-misuse-and-blocking-high-risk-activity"},{"level":2,"text":"Defending critical infrastructure","id":"defending-critical-infrastructure"},{"level":2,"text":"Expanding access to trusted defenders","id":"expanding-access-to-trusted-defenders"}]}}