How Check Point accelerates vulnerability research with Daybreak

Check Point, which says it helps secure more than 100,000 organizations, is using OpenAI's Daybreak to speed the research that turns a newly disclosed vulnerability into a customer protection. Its CTO, Jonathan Zanger, says work that took hours now takes minutes.

In security, that gap is the whole contest. The faster defenders can understand a flaw and ship a signature or indicator of compromise, the smaller the window attackers get. If AI can compress that window, that is one of the clearest wins for useful AI we've seen.

We're broadly in favor of this. Vulnerability research has been bottlenecked by the number of skilled analysts available, and Check Point says as much. A model that can understand a vulnerability, reason through attack paths and scan code for gaps doesn't replace those analysts; it multiplies them. Getting protections out in minutes instead of hours is a real public good when the threat is spreading.

But cyber is the most dual-use domain in AI. The same capabilities that let a defender reason through how a system could be exposed are, by definition, capabilities an attacker would want. That's why the structure matters. Daybreak is offered to defenders through a defense network, not as a general-purpose tool, and that's the right default. The question is how well that gate holds as more partners join and the models get stronger.

The next step is where we'd slow down and pay attention. Check Point is working with OpenAI on an agentic network security orchestration system that would continuously monitor customer environments and recommend actions. Zanger describes clear operating boundaries and human approval for consequential actions, with the system explaining its findings, recommended actions and risks. Those are exactly the right words. We'll be watching whether they survive contact with a busy security team at 2 a.m., when approving everything becomes the path of least resistance.

There's also a quieter risk: speed of detection is not speed of remediation, and Check Point itself names remediation as its next priority. Finding gaps faster than organizations can fix them can create alert fatigue and a backlog of known weaknesses. If this scales, the winners will be teams that pair faster research with faster patching and clear accountability. The losers will be anyone who treats automated security as a reason to staff down.

One caveat on sourcing: this is a customer story from the model provider, built on the CTO's account. The hours-to-minutes claim is plausible, but it isn't independently measured here.

Check Point Research investigates emerging vulnerabilities and turns findings into protections across the company's AI, workspace, cloud and network security products. Daybreak gives participants in OpenAI's Daybreak Defense Network access to frontier cyber models for analyzing vulnerabilities and security gaps in code and configurations.