- OpenAI has unveiled Dawn, its newest safety challenge
- It seeks to rival Anthropic’s Mythos in detecting and patching high-severity vulnerabilties
- Dawn can even assist corporations construct software program extra securely from the beginning
OpenAI has unveiled Dawn, its reply to Anthropic’s Mythos and Challenge Glasswing, sparking a possible cybersecurity arms race between the 2 corporations.
“Dawn combines the intelligence of OpenAI fashions, the extensibility of Codex as an agentic harness, and our companions throughout the safety flywheel to assist make the world safer for everybody,” the announcement stated.
The challenge seems to work with OpenAI’s trade and authorities companions by securing software program on the very starting of the event course of, making a extra strong base that, in time, will scale the effectiveness of cyber protection.
Dawn to construct software program securely from the beginning
Within the weblog put up, OpenAI sums up the challenge in a single sentence: “The objective is straightforward: speed up cyber defenders and repeatedly safe software program.”
Dawn will permit organizations to use OpenAI’s Codex Safety to their very own repository utilizing an ‘agentic harness’, the place it should hunt down, analyze, and patch assault paths and high-impact code.
Excessive-priority vulnerabilities may be analyzed and validated in a safe, remoted surroundings, “so groups can prioritize actual, reproducible points over noisy alerts.” Codex Safety can even permit groups to automate detection and response, growing effectivity and securing important vulnerabilities sooner.
Dawn due to this fact seeks to delegate the rote work of figuring out and analyzing to AI, and to return the evidence-backed outcomes of vulnerabilities to human groups. The place Dawn differs in its strategy in comparison with Mythos is in constructing software program securely from the beginning and continually monitoring for vulnerabilities, in comparison with Mythos’ focus of detecting and mitigating high-severity vulnerabilities at scale.
Dawn contains three fashions; GPT-5.5, because the default for general-purpose work; GPT-5.5 with Trusted Entry for Cyber, for use in defensive safety workflows; and GPT-5.5-Cyber, for specialised workflows together with purple teaming and pen testing.
Dane Knecht, CTO at Cloudflare, stated, “We’re excited in regards to the potential of OpenAI’s cyber capabilities to carry stronger reasoning and extra agentic execution into safety workflows. It’s a giant step ahead for groups to have the ability to leverage frontier fashions not solely to speed up velocity, but additionally to enhance their safety posture.”
Comply with TechRadar on Google Information and add us as a most popular supply to get our knowledgeable information, evaluations, and opinion in your feeds.
