After Anthropic’s Mythos, OpenAI Brings Daybreak To Find Software Bugs

Sam

Altman-led OpenAI has introduced Daybreak, a new artificial tool designed to identify, detect and fix critical software flaws faster. The company said this AI system will be powered by the GPT-5.5 and focus on defensive cybersecurity work. This announcement seems to be an answer to Anthropic’s Claude Mythos AI which gained attention worldwide for its capabilities to identify hidden software bugs before even hackers can. Although it has been making headlines for its same ability which can also be used by hackers for potential misuse.

How OpenAI’s Daybreak Works?

OpenAI revealed that its new AI system improves how critical software vulnerabilities can be identified and helps to spot early risks like review code and test patches to bring quick fixes. Moreover, the company added that this tool can help experts to examine large codebases, identify hidden bugs, analyse unfamiliar systems and speed up the process of fixing security problems. It is noteworthy that just like Anthropic, even OpenAI acknowledges in its blog post that the AI capabilities of these tools could also be misused. The American tech firm stated that to avoid such issues, it also includes safeguards, verification systems, monitoring tools and account-level controls for critical cybersecurity tasks.

Who Can Use Daybreak

The company has rolled out different tiers for cybersecurity-related AI usage. Notably, the default GPT-5.5 system will offer standard measures for general developer and knowledge work tasks. Another version known as ‘GPT-5.5 with Trusted Access for Cyber’ is created for verified defensive cybersecurity work. The company noted that this iteration can be utilised for secure code reviews, malware analysis, vulnerability triage, detection engineering and patch validation in authorised environments. Additionally, the tech firm has also introduced ‘GPT-5.5-Cbyer’ that offers wider permissions for specialised workflows such as authorised penetration testing and red teaming. The company notes that this iteration can include much stronger verification requirements and strict account-level controls. What lies ahead is that OpenAI revealed it plans to work with government partners and industries in the coming weeks as it prepares wider deployment. However, it has yet to be seen how Cyberbreak and Mythos function in future as they have the capabilities that may potentially wreak havoc especially for the banking industries as per several cybersecurity experts.