SentinelOne (S) says OpenAI's Daybreak tool can produce human-level malware reverse engineering in roughly eight hours, a job that can take a top researcher four to six weeks. If that holds up, security teams could look at far more threats than they can today.
The claim comes from Gregor Stewart, SentinelOne's Chief AI Officer, in a case study published by OpenAI. That matters for how you read it: this is a customer story on the vendor's own site, not an independent test. We haven't found any outside researcher who has checked the eight-hour figure.
What Daybreak does for the team
Daybreak gives defenders access to OpenAI models for cybersecurity work. SentinelOne uses it to reverse engineer malware, meaning it pulls software apart to understand how it behaves.
The real gain, Stewart says, is capacity. Deep investigation used to be reserved for the most important threats. "We were only able to reverse very small numbers of pieces of software," he says. With more samples analyzed, researchers can spot shared components, such as a piece of code found in two seemingly unrelated samples that could become a lead in a customer investigation.
Humans still own the decisions
Stewart is careful about automation. "The only thing that can't be automated is accountability," he says. Architecture changes should still get human eyes, while point fixes may suit automation.
What the post leaves out matters too. As far as we can tell from the case study, it doesn't explain how "human-level" was measured, how many samples have been processed, or whether researchers still review output line by line. It also gives no pricing, and it doesn't describe a single case where an analyst disagreed with or changed what the AI produced. Those details decide whether eight hours is typical or a best case, and whether the human oversight is real or a talking point.
We haven't been able to get answers on any of this. Questions to SentinelOne and OpenAI about methodology, sample counts, pricing and examples of human overrides, and calls to independent security researchers, are still outstanding, so treat the eight-hour figure as a company claim until someone outside checks it.
Next, SentinelOne wants to bring OpenAI's cyber capabilities into code creation, so developers can catch security problems as they build. That is a plan, not a shipped feature.
If an AI can do a month of expert work in a day, who checks the checker?
Sources
Reported by the WattsUpNext desk from the sources linked below. Spot an error? Tell us at corrections@wattsupnext.com.
The WattsUpNext Brief
Get stories like this in your inbox.
One email each weekday, only the topics you choose. Real news, sourced, no fluff. Unsubscribe in one click.
