OpenAI Expands Daybreak Cybersecurity Program With Red-Team Model and 16 Partners
Summary
OpenAI expanded Daybreak with models for defensive security and red-teaming, and announced 16 provider partnerships. In a company-devised evaluation, it said GPT-5.6-Cyber completed 95% of complex hacking tasks versus 1.5% for Sol.
Key points
- OpenAI expanded its Daybreak program, which provides unreleased models to private organizations and governments for defensive cybersecurity work.
- Daybreak Blue supports vulnerability discovery, secure code review, malware analysis, incident response, and patch validation, with reduced cybersecurity safeguards.
- Daybreak Red offers GPT-5.6-Cyber for advanced red-teaming and exploit testing; OpenAI said it is less likely to refuse dual-use cyber requests.
- OpenAI said Daybreak Red use will be closely monitored and supervised because GPT-5.6-Cyber is more capable at malicious cyber tasks than Sol.
- In an evaluation designed by OpenAI, the company reported that GPT-5.6-Cyber completed 95% of complex hacking requests, compared with 1.5% for Sol.
- OpenAI announced partnerships with 16 cybersecurity providers to make its models available through existing security services and help defenders find, validate, and fix vulnerabilities.
- The article notes that experts say AI systems still need substantial human guidance and supporting infrastructure, and may struggle to patch vulnerabilities safely without such guidance.
Article Details
- Event Type
- Expansion of OpenAI’s Daybreak cybersecurity program and launch of a cybersecurity provider partnership program
- Impact
- OpenAI said Daybreak Blue supports defensive work including vulnerability discovery, secure code review, malware analysis, incident response, and patch validation. Daybreak Red provides access to GPT-5.6-Cyber for advanced red-teaming and exploit testing, under close monitoring and supervision. In OpenAI’s evaluation, GPT-5.6-Cyber completed 95% of complex hacking requests, compared with 1.5% for ChatGPT-5.6-Sol. The expanded capabilities could help defenders identify and fix vulnerabilities faster, while OpenAI acknowledged risks from misuse or model misalignment.
Vendors
AccentureThe partners include IBM, CrowdStrike, Accenture, Ernst & Young, KPMG, Palo Alto Networks, Cisco, Cloudflare, Sophos and others.Ciscoprogram with 16 major cybersecurity providers, including IBM, CrowdStrike, Palo Alto Networks, Cisco, Cloudflare, and Sophos, to help defenders find, validate, and fix serious vulnerabilities faster.Cloudflareprogram with 16 major cybersecurity providers, including IBM, CrowdStrike, Palo Alto Networks, Cisco, Cloudflare, and Sophos, to help defenders find, validate, and fix serious vulnerabilities faster.CrowdStrikeThe company also launched a partnership program with 16 major cybersecurity providers, including IBM, CrowdStrike, Palo Alto Networks, Cisco, Cloudflare, and Sophos, to help defenders find, validate, and fix seriousErnst & YoungThe partners include IBM, CrowdStrike, Accenture, Ernst & Young, KPMG, Palo Alto Networks, Cisco, Cloudflare, Sophos and others.IBMThe company also launched a partnership program with 16 major cybersecurity providers, including IBM, CrowdStrike, Palo Alto Networks, Cisco, Cloudflare, and Sophos, to help defenders find, validate, and fix seriousKPMGThe partners include IBM, CrowdStrike, Accenture, Ernst & Young, KPMG, Palo Alto Networks, Cisco, Cloudflare, Sophos and others.OpenAIOpenAI expanded its Daybreak program with new frontier model variants for defensive cybersecurity and advanced red-teaming.Palo Alto Networkslaunched a partnership program with 16 major cybersecurity providers, including IBM, CrowdStrike, Palo Alto Networks, Cisco, Cloudflare, and Sophos, to help defenders find, validate, and fix serious vulnerabilitiesSophos16 major cybersecurity providers, including IBM, CrowdStrike, Palo Alto Networks, Cisco, Cloudflare, and Sophos, to help defenders find, validate, and fix serious vulnerabilities faster.
Products
AstraLast week, OpenAI said it was intentionally slowing down development of its newer “Astra” model in order to develop better guardrails to restrain its behavior.ChatGPT-5.6-SolDaybreak Blue, powered by OpenAI’s ChatGPT-5.6-Sol, would operate with lower cybersecurity safeguards compared to other commercially available models and is described as “a recommended starting point for most defenders”Daybreak BlueDaybreak Blue supports vulnerability discovery, secure code review, malware analysis, incident response, and patch validation.Daybreak RedDaybreak Red uses GPT-5.6-Cyber for advanced red-teaming and exploit testing.GPT-5.6-CyberDaybreak Red uses GPT-5.6-Cyber for advanced red-teaming and exploit testing.