OpenAI announces Daybreak, its answer to Claude Mythos

OpenAI has launched Daybreak, its most advanced AI initiative for cybersecurity. This move comes just weeks after Anthropic announced Claude Mythos, an AI model that has raised alarms over global cybersecurity. Here is how OpenAI's cybersecurity project works.

In Short

• OpenAI announces Daybreak, its cybersecurity AI project
• This is likely OpenAI’s answer to Anthropic’s Claude Mythos
• OpenAI says it is willing to work with many companies with Daybreak

Last month, Anthropic shocked the global cybersecurity world with Claude Mythos – an AI so powerful that it could potentially hack any software or operating system. Companies and governments around the world, including India, raised concerns over the future of security online due to Mythos. And now, OpenAI may finally have its own Claude Mythos, in the form of Daybreak.

OpenAI CEO Sam Altman announced Daybreak on X. He stated, “OpenAI is launching Daybreak, our effort to accelerate cyber defense and continuously secure software.”

Altman insisted that OpenAI was willing to “start working with as many companies as possible” with Daybreak to help them out when it comes to cybersecurity. Do note that Anthropic has made Claude Mythos available only to a selected group of around 40 companies, such as Amazon and Google, via Project Glasswing.

What is OpenAI Daybreak?

In its announcement, OpenAI said Daybreak is based on the idea that the next era of cyberdefense should be resilient by design. That is, companies need to work on building stronger cyberdefenses at a time when AI is only getting better.

While AI can help defenders when it comes to identifying vulnerabilities and validating fixes, these capabilities could also be misused by bad actors.

As per OpenAI, Daybreak combines the company’s existing AI models and Codex’s agentic capabilities. Codex is OpenAI’s coding agent that has become widely popular in recent weeks.

The Sam Altman-led firm says that defenders can use Daybreak for a host of different tasks related to cybersecurity. This includes secure code review, threat modelling, patch validation, and dependency risk analysis.

Is it similar to Claude Mythos?

On the surface, it may seem that OpenAI’s Daybreak is quite similar to Anthropic’s Claude Mythos. However, there are some key differences.

While Anthropic’s Project Glasswing uses solely Claude Mythos, an unreleased AI model, OpenAI’s Daybreak uses different AI models such as GPT-5.5 and GPT-5.5-Cyber.

Another difference here is that OpenAI is willing to work with many companies, as per Sam Altman. In its blog post, the AI startup stated that it was already working with the likes of Cloudflare, Oracle and Cisco.

On the other hand, Anthropic has largely restricted Mythos to around 40 companies via Project Glasswing. Though a group of unauthorised users did allegedly manage to get their hands on this model.

How does Daybreak work?

OpenAI states that Daybreak is meant to help teams find and fix vulnerabilities, burn down the backlog, automate detection and response, and choose the right level of access for different workflows.

OpenAI listed three model options under Daybreak. GPT-5.5 is the default model with standard safeguards for general-purpose, developer and knowledge work. GPT-5.5 with Trusted Access for Cyber is meant for verified defensive work in authorised environments. This includes secure code review, vulnerability triage, malware analysis, detection engineering and patch validation.

GPT-5.5-Cyber is described as the most advanced option for specialised authorised workflows, paired with stronger verification and account-level controls, with preview access for authorised red teaming, penetration testing and controlled validation.

The company said Daybreak builds on its earlier cyber work. One of the descriptions of the launch said the initiative builds on OpenAI's April release of GPT-5.4-Cyber, which the company said had contributed to fixing more than 3,000 vulnerabilities.

- Ends