We partnered with @fmf_org and its member firms on a new framework to help identify when bypassing an AI system's safeguards could enable serious cyber harm, and prioritize stronger protections.

"...This framework is an important step toward a common methodology for evaluating and comparing the severity of cyber jailbreaks, including the capabilities they unlock and how readily they can be used. Shared approaches like this can help the AI and security communities build a clearer evidence base and prioritize improvements to safeguards." - OpenAI's Sasha Baker, Head of National Security Policy

frontiermodelforum.org/updates/introd…

A new way to measure how AI agents handle the routine analyses of wet-lab science—BioBasicsBench.

Developed with Ligo, a biotech company focused on enzyme design, the benchmark draws on 100 tasks from real experiments to test whether agents can interpret raw lab data, identify hen the evidence doesn’t support a conclusion.

Accelerating biological research starts with getting the everyday work right. BioBasicsBench shows where models need to improve, and helps us track what those gains mean for scientific progress, biosafety, and biosecurity.

openai.com/index/measurin…

On November 12, also join our Cyber at the Frontier livestream at 10am PT for more Cyber announcements: openai.com/cyber-at-the-f…

In this narrow defender’s window, our goal is to expand Daybreak access to organizations of every size, independent researchers, and open-source maintainers.

Daybreak Blue helps verified defenders validate vulnerabilities and fixes with fewer refusals. Daybreak Red enables offensive security with deeper exploit development and authorized testing.

Today we are also expanding the Daybreak Blue alpha for GPT‑6 Astra to more organizations and individual researchers.

Sign up for Daybreak: openai.com/business/solut…

Introducing GPT‑6 Cyber: a major advance in security research that pairs stronger capabilities with better model alignment.

The model powered the Linux kernel research we announced yesterday, and security teams at companies like Mozilla, Palo Alto Networks, Calif, and at OpenAI—are already putting it to work.

GPT-6 Cyber is available to select Daybreak Red customers and we are expanding access over time.

openai.com/index/introduc…

Images 2.6 is available today in ChatGPT, and Codex across all tiers on web, desktop, and mobile.

GPT-Image-2.6 Flare and Sunburst are also available in the API.

openai.com/index/introduc…

Two new image models for developers: GPT-Image-2.6 Flare and Sunburst.

Flare generates images in as little as six seconds. Sunburst brings higher-quality generation and greater consistency across edits.

Image markup and SVG exports give you more ways to refine and use what you create.

Save once. Create again and again.

Keep reference images of yourself, pets, and characters ready to reuse in new creations—no need to upload them again.

GIFS!

With ChatGPT Images 2.6, turn a prompt or image into an animated GIF.

Adjust the speed, edit individual frames, and share it on repeat.

Images, upgraded—introducing ChatGPT Images 2.6.

Now with GIFs and reference images.

Plus, two new models are coming to the API with 50% faster image generation.

Our goal is to help defenders find and fix vulnerabilities before attackers can use them. We’re working with Linux maintainers and Patch the Planet partners on reports and patches.

openai.com/index/a-thousa…

The agents also discovered and used at least 10 Linux exploitation techniques that, to the best of our knowledge and all of the experts that we consulted, are novel.

This suggests frontier LLMs can discover and patch new approaches to exploitation, beyond simply applying known techniques.

The full run cost $1,213,430 in API credits—about $1,279 per distinct local privilege escalation vulnerability.

For context, applying historical Pwn2Own and TyphoonPWN bounty rates gives a hypothetical value of $18.9M–$66.3M.

This demonstrates the relative efficacy of using AI models for securing software—defenders should proactively analyze software to make cyber attacks cost-prohibitive.

The run reproduced 1,724 vulnerabilities. It produced proofs of concept for 291 information leaks and working exploits for 948 distinct local privilege escalation vulnerabilities.

Those exploits worked on a stock Linux desktop configuration with standard security protections enabled.

The pipeline finds security-relevant code, scans for vulnerabilities, reproduces them in isolated environments, tests whether they can be exploited, and generates candidate patches.

Human researchers review the results before reporting vulnerabilities and propose fixes to maintainers.

Our security researchers built an autonomous cyber vulnerability discovery, exploitation, and patching pipeline.

In a single run with an upcoming cyber model, it validated and generated patches for 948 local privilege escalation vulnerabilities in the Linux kernel at a total API cost of $1.2 million.

These are early steps, and we’re looking for engineers, engineering firms, and software partners to help shape what comes next.

openai.com/index/accelera…

We’re also working with engineers at Chip Ganassi Racing to turn 3D scans into editable CAD models and manufacturing drawings, and with Mazda to explore new vehicle designs through simulation.

Engineers review the outputs and help identify where the models need to improve.

Two OpenAI team members worked with GPT-6 Astra to design a prototype device for talking to Codex in under a week, as a side project.

Astra designed the enclosure, circuit board, interface, and software. Getting it manufactured still required several rounds of review and iteration.

ChatGPT and Codex are helping engineers build for the physical world.

We’re sharing our progress across mechanical design, electrical engineering, and simulation—from helping our hardware teams iterate faster to testing new workflows with industry collaborators.