OpenAI has expanded its Daybreak cybersecurity program
On August 10, OpenAI announced the expansion of its closed-door cybersecurity initiative, Daybreak. The program will have two access tiers—Daybreak Blue and Daybreak Red—to provide participants with tools to defend against cyberattacks, according to CNBC.
OpenAI launched Daybreak in May for partners in its ecosystem, enabling them to apply state-of-the-art artificial intelligence models amid the rapidly changing cyberthreat landscape. The company stated that it aims to provide cutting-edge AI capabilities to trusted defenders before attackers can deploy AI on a large scale for offensive operations.
Daybreak Blue participants will receive special access to OpenAI’s cutting-edge general-purpose models. The company will adjust the models’ security restrictions to allow for the execution of defensive cybersecurity tasks. OpenAI recommends this tier as the starting point for most organizations.
For more breaking news, follow the UA.News Telegram channel.
The Daybreak Red tier provides access to specially trained cybersecurity models. These can be used to test system security, research vulnerabilities, and validate exploits. These participants will also gain access to the new GPT-5.6-Cyber model, based on GPT-5.6 Sol. According to OpenAI, it was developed to improve performance on specialized cybersecurity tasks and reduce the number of rejections for such requests.
The Daybreak expansion followed several cybersecurity incidents reported in recent weeks by AI developers, including OpenAI, Anthropic, and Meta. According to CNBC, in each case, an AI model gained access to systems that were supposed to remain inaccessible during cybersecurity testing. This prompted industry researchers and officials to call for stronger security measures.
Last week, OpenAI also announced the suspension of some internal work related to its upcoming Astra model. During testing, the model demonstrated significant progress in agent-based programming and cybersecurity, so the company is evaluating these capabilities and working to strengthen security measures and controls.
