EXPLORINEER
en

KR — 전자신문

OpenAI Releases 'GPT-6 Cyber,' a Cybersecurity-Focused AI Model

AI2 min read

OpenAI has released GPT-6 Cyber, an AI model built to find and help fix security flaws in software. It is first available only to approved security researchers and customers.

What happened

On the 6th, OpenAI released GPT-6 Cyber, a version of its AI made for cybersecurity. The model helps check software for weak spots and helps build fixes called security patches. For now, only some approved customers in the 'Daybreak Red' program can use it, through an API (a way for programs to talk to the AI) and Codex. OpenAI plans to open it to all Daybreak Red customers later. The AI finds security flaws in code, checks whether attackers could use them, and writes possible patches and analysis notes. It was also trained to follow the work scope and limits that researchers approve.

Key numbers

  • OpenAI's researchers used GPT-6 Cyber to find 81 high-risk flaws in Mozilla's Firefox browser.
  • In Linux kernel research, the model reproduced 1,724 flaws. It built working attack code (exploits) for 948 of them.
  • 304 main findings were reported to software maintainers, along with possible patches.
  • On the ExploitGym test, which measures how well an AI can use flaws, GPT-6 Cyber scored 48.5%. The previous model, GPT-5.6 Cyber, scored 32.1%, a gap of 16.4 points.
  • In an early test of a realistic server setup (RemoteExploitBench), it finished 95% of tasks.

Background

Security experts often say that AI can help defenders find and fix bugs faster than before. The same skills, however, could help attackers. That is why companies try to limit who can use powerful security tools and set rules about what they can test. OpenAI also said it added safeguards. These include blocking high-risk cyber actions, detecting misuse, and monitoring whether the model stays within approved tasks. Requests that could harm critical infrastructure, such as power or water systems, are limited, while approved security research is still supported.

Why it matters

Software flaws can cause big problems, from stolen data to broken services. If AI can find and fix flaws faster, safer software could follow. At the same time, a tool that can write attack code is risky if it is misused. The results show both the promise and the need for strict limits. How well OpenAI controls access will shape whether this helps defenders more than attackers.

What to watch

  • Whether OpenAI opens GPT-6 Cyber to all Daybreak Red customers as planned.
  • Whether the 304 reported flaws are fixed by software maintainers.
  • How the safeguards hold up in real use outside of tests.
  • How rival AI companies respond with their own security models.

Glossary

API
A tool that lets other programs send requests to an AI model and get answers back.

Ad

Add 150 words. Learn 10 a day.

Dicter word books send only the words you learn to your review notes.

More in this topic

AIITmedia

Fuji TV to Use Generative AI in All Dramas by 2027; Up to 80% Labor Cut in Some Production Steps

Fuji Television announced on November 5 that it will have a system in place by October 2027 to use generative AI in all dramas it produces. Starting with the October 2026 season, at least one drama per quarter will be designated an "AI Strengthened Slot," and the company says some video production steps have already cut labor by up to 70–80% compared with conventional methods, though this applies only to specific steps, not whole-production costs.

AIThe Verge

OpenAI is stepping back from NDAs for data center projects

After this week's midterms, in which voters voiced opposition to data center expansion, OpenAI announced new commitments to communities where it works on data center projects. These include not using non-disclosure agreements with public officials, paying the full cost of the electricity and infrastructure it needs, and publishing expected jobs and hiring reports. OpenAI also said future "compute campuses" it owns and operates will have a "local compact" reflecting local priorities.

AIITmedia

OpenAI Unveils Cybersecurity-Focused "GPT-6 Cyber," Finds 81 Unknown Firefox Vulnerabilities

OpenAI announced GPT-6 Cyber, a cybersecurity-specialized AI model built on the general-purpose GPT-6 Astra, and will initially offer it only to approved users of its Daybreak Red defender program. In research on real software, the model found 81 unknown high-severity vulnerabilities in Firefox and developed working exploits for 948 Linux kernel vulnerabilities, and OpenAI says it is also restricting access for critical infrastructure organizations.

AIThe Verge

Perplexity launches its own take on AI agents for work

Perplexity has launched "Teammates," always-on AI agents for work that can take on multiple roles within a workspace and draw on existing projects, skills, and memory from Perplexity tools. Teammates integrate with Perplexity Computer, which deploys agents on a user's device, and each Teammate gets its own thread that can be tracked in the Perplexity app or accessed in Slack. The feature requires Perplexity's $200-per-month Max plan.