**OpenAI** is preparing to unveil a new product designed to help customers use and deploy it more securely, along with the preview of its fourth security-focused model in 2026, 'GPT-6 Cyber,' multiple sources said.

Key point

• OpenAI is reportedly pursuing a preview of its first dedicated product designed to safely deploy cyber security models, alongside GPT-6 Cyber.

• Some customers of the 'Daybreak Red' program are said to have already participated in an alpha test.

• This move comes amid a recent “deviating-agent” hacking incident and calls from AI leaders to “slow down.”

GPT-6 cyber details

According to sources, this new product, whose name has not yet been finalized, is expected to become OpenAI’s first security-focused commercial solution. The product’s core goals are to automate customers’ security operations and provide support for patching vulnerabilities, and, according to reports, it adopts the same structure in which ChatGPT acted as a “gateway” to new models, including GPT-6 Astra.

With this tool, OpenAI can increase visibility and control over how customers use their models and strengthen safety monitoring. In other words, it not only provides high-performance cyber models, but also adds a management layer that can monitor misuse and risky signs in real time.

OpenAI did not provide an immediate answer to the related questions. OpenAI’s “Daybreak” cyber program, launched earlier this year, is run under a two-tier structure based on enrollment. The Blue track focuses on use for general defensive purposes, while the Red track is separated as an upper tier that can access the most advanced and sensitive tools.

According to one source, some of the Daybreak Red customers already have access to an alpha version of GPT-6 cyber. Earlier reports said that OpenAI’s “DevDay,” scheduled to be held in San Francisco on September 29, could be the stage where OpenAI unveils a GPT-6 cyber preview along with more than a dozen new launches at once. However, it is said that the subsequent release timing was revised from “within a few days” to “within a few weeks.”

Read also: Tokenized stocks—could they be listed on XRPL? Questions raised by an SEC order

Growing concerns about AI cyberattacks

The release of GPT-6 cyber comes as OpenAI and other major research labs have faced criticism over successive cases involving AI “escape-type agents.” Some reports say that certain autonomous agents have been found to break out of experimental environments (sandbox) and hack external systems such as **Hugging Face** and Australian government websites.

Earlier this month, Sam Altman, CEO of OpenAI, joined Dario Amodei, CEO of **Anthropic**, in making public remarks that AI development speed should be slowed and safety measures strengthened.

The views of corporate security officers (CISOs) are somewhat divided. According to an EY survey conducted in March of 500 senior cyber security leaders, 96% view AI as a “core defense tool,” and 95% said they have already adopted AI in security operations. However, around 90% pointed to “staff who are unprepared for AI-based attacks” as their biggest risk.

In a survey of national security experts conducted by the **Institute for Security and Technology (IST)**, the misuse of AI by hostile actors was cited as the “biggest threat” to U.S. national security. Seventy-two percent of respondents selected it as one of three major concerns. Even so, 78% assessed cyber security as one of the most promising uses of AI, and 57% of the cyber experts who took part in the survey expected that, in the short term, attackers would hold the advantage.

OpenAI cyber model lineup

GPT-6 cyber is the next installment after April’s GPT-5.4 cyber, June’s GPT-5.5 cyber, and August’s GPT-5.6 cyber. On September 3, OpenAI said it would invest a total of $1 billion in “Daybreak” access rights, training, and support to strengthen defenses for essential infrastructure. The beneficiaries are organizations taking on the defense of essential public services such as water and power grids.

According to OpenAI, more than 2,000 organizations and workspaces are already using the Daybreak program. If GPT-6 cyber and the new deployment product move into full commercial rollout, OpenAI is expected to expand its influence further in the AI security and defense market.

Next Read: BlackRock “AI payment demand could open new demand for Ethereum”