OpenAI has launched Dawn, a brand new “frontier AI for cyber defenders” initiative targeted on serving to organizations construct safer software program that’s resilient by design. The corporate describes Dawn as its imaginative and prescient for altering how software program is constructed and defended by integrating cyber protection straight into the software program improvement lifecycle.
OpenAI says Dawn represents “the primary glimpse of daylight within the morning.” For cyber protection, the initiative is designed round seeing dangers earlier, appearing sooner, and serving to make software program resilient from the beginning.
In line with the corporate, the following period of cyber protection ought to transfer past merely discovering and patching vulnerabilities. As an alternative, software program methods ought to be designed to stay resilient even when vulnerabilities exist.
OpenAI says AI may help defenders motive throughout massive codebases, establish delicate vulnerabilities, validate fixes, analyze unfamiliar methods, and transfer from discovery to remediation quicker. The corporate additionally says these capabilities are paired with belief, verification, proportional safeguards, and accountability to cut back misuse dangers.
The said purpose of Dawn is to “speed up cyber defenders and repeatedly safe software program.”
Key Dawn options and capabilities
Dawn combines OpenAI fashions, the extensibility of Codex as an agentic harness, and companions throughout what the corporate calls its “safety flywheel” to help AI-powered cyber protection workflows.
Key capabilities embrace:
- Safe code evaluation
- Menace modeling
- Patch validation
- Dependency danger evaluation
- Vulnerability detection
- Remediation steerage
- Malware evaluation
- Detection engineering
- Automated monitoring and response workflows
OpenAI says these capabilities may be built-in into on a regular basis improvement environments so software program turns into extra resilient from the start of the event cycle.
The corporate outlined a number of methods organizations can deploy AI in cyber protection workflows:
- Prioritize high-impact threats and scale back investigation time from hours to minutes
- Generate and check patches straight inside repositories
- Confirm remediation utilizing audit-ready proof and validation methods
- Construct editable risk fashions from repositories
- Focus evaluation on life like assault paths and high-impact code areas
- Validate vulnerabilities in remoted environments to cut back noisy alerts
- Assist organizations prioritize reproducible points and “burn down the backlog” quicker
- Automate monitoring, detection, and response for higher-risk vulnerabilities
OpenAI additionally confirmed that it’s working with authorities and business companions forward of deploying more and more cyber-capable AI fashions by way of its iterative deployment method within the coming weeks.
GPT-5.5 entry tiers for cyber protection
OpenAI launched a number of entry ranges for organizations relying on safety workflow necessities.
GPT-5.5 (default)
- Commonplace safeguards for general-purpose use
- Supposed for developer, productiveness, and basic data work workflows
GPT-5.5 with Trusted Entry for Cyber
- Extra exact safeguards for verified defensive work in approved environments
- Designed for safe code evaluation, vulnerability triage, malware evaluation, detection engineering, and patch validation
GPT-5.5-Cyber
- Most permissive habits for specialised approved workflows
- Contains stronger verification methods and account-level controls
- Supposed for preview entry workflows together with approved pink teaming, penetration testing, and managed validation
OpenAI says organizations can contact the corporate to find out essentially the most appropriate entry mannequin for his or her safety workflows.
Availability
OpenAI is at the moment allowing organizations to request vulnerability scans and Dawn assessments to establish, validate, and remediate safety points throughout functions and codebases.
The corporate says organizations can begin with a Dawn evaluation to see how AI may help prioritize danger, remediate points quicker, and strengthen software program defenses.
Companies concerned about this system are required to submit firm particulars, group measurement, contact data, work e-mail addresses, and cellphone numbers by way of OpenAI’s request kind.