OpenAI launches GPT-5.6-Cyber, its new advanced cybersecurity model, and expands the Daybreak program

OpenAI has announced the GPT-5.6-Cyber ​​release its new model specialized in advanced cybersecurity, and the expansion of Daybreak its global cybersecurity program for the detection and correction of vulnerabilities for organizations, which will now be available with access levels Daybreak Blue y Daybreak Red.

GPT-5.6-Cyber ​​has been developed based on GPT-5.6 Sol launched last month, although with some differences focused on its specialization for advanced security research tasks, vulnerability development and attack testing.

In fact, as the Artificial Intelligence (AI) laboratory has explained on its website, this new model has been tested with ‘benchmarks’ like ExploitGym which aims to evaluate whether agents can convert vulnerabilities into functional exploits. In the results, GPT-5.6-Cyber ​​outperformed both GPT-5.6 Sol and GPT-5.5-Cyber.

Following this line, one of the areas where OpenAI has strengthened GPT-5.6-Cyber It is in the ability to find and accurately calibrate the severity of new zero-day vulnerabilities, that is, a security flaw unknown to developers and with no solution available when discovered.

Also, compared to GPT-5.6 Sol, GPT-5.6-Cyber completes 95 percent of advanced and dual-use cybersecurity requestswhile Sol rejects almost all of them, completing only 1.5 percent.

For example, OpenAI researchers have claimed that they used GPT-5.6-Cyber ​​to identify previously unknown vulnerabilities in V8Chrome’s JavaScript engine, which could be chained together to corrupt memory and escape the heap sandbox.

The technology company has also clarified that, when it launched its GPT-5.6 Sol frontier model, it included a series of cybersecurity safeguards. These safeguards now they wake up with the Daybreak program which expands its availability to more organizations.

TWO LEVELS OF ACCESS TO DAYBREAK: BLUE AND RED

Specifically, the technology company has expanded access to Daybreak dividing it into two modalities with different verification requirements.

On the one hand, it has announced Daybreak Bluea level of access adjusted for typical defensive work that companies or organizations may require, such as code review or malware analysis. This level provides access to general models such as GPT-5.6 Sol, although removing the restrictions that apply to the models available to all users.

On the other hand, it has launched the level Daybreak Red, offering advanced access to cutting-edge models for researchers and test teams, this includes the new specialized GPT-5.6-Cyber ​​model.

However, OpenAI has clarified that access for these levels is limited to “approved individuals and organizations” carrying out authorized work, with identity verification, account security, monitoring, approved use restrictions and legal declarations.

In this sense, OpenAI is expanding Daybreak through security and services partners, including Accenture, Cisco, Cloudflare, CrowdStrike, IBM, Unit 42 of Palo Alto Networks and others. “This allows more organizations to have a way to use cutting-edge cybersecurity models through trusted products, managed services and collaborations with clients,” he added.

Through the program, OpenAI has influenced that community initiatives are created, in reference to alliances with security companies, as well as with the open source community, and AI-assisted review is promoted to find and correct bugs.