SentinelOne reports that OpenAI's Daybreak model can reverse-engineer sophisticated malware in about eight hours, a task that previously took top security researchers four to six weeks. Gregor Stewart, SentinelOne's chief AI officer, says the model produces output at a human level, freeing researchers to investigate more threats than before.
The time savings allow SentinelOne to analyze a larger portion of the malware landscape and compare samples for shared components. Stewart says this broader view reveals relationships between threats that isolated investigations might miss, helping researchers defend more proactively.
Stewart emphasizes that accountability cannot be automated. While point fixes may be suitable for automation, decisions that change software architecture still require human review. He also notes that access to OpenAI's cyber capabilities is spreading security expertise across more team members, and he hopes to bring these capabilities into code creation so developers can address security issues as they build.