Juan Andrés Guerrero-Saade argues that cybersecurity is leaving its experimental era behind, as non-standardized complexity has made systems too costly and difficult to manage through human attention alone. He says large language models can provide scalable evaluative power and help defense become more standardized, automated, and sustainable, with #SentinelLABS #LABScon #OpenAI #JAGS.
Keypoints
- Juan Andrés Guerrero-Saade (JAGS) delivered the keynote in the final LABScon 25 Replay video.
- He argued that cybersecurity has reached the end of its experimental era because complexity has outpaced human-only management.
- JAGS described large language models as a source of cheap, effectively unlimited evaluative power.
- He framed LLMs as a “lossy compression of human knowledge” that can help defenders assess, prioritize, and act at scale.
- The talk emphasized that AI should be built into security systems, not merely bolted on afterward.
- Drawing on cybernetics, he called for security designs that combine human expertise with artificial evaluative power.
- The broader vision is a more standardized, automated, and sustainable future for cybersecurity workflows and products.
MITRE Techniques
Indicators of Compromise
- [Organizations/Events ] conference and research venue context – SentinelLABS, LABScon 2025, and LABScon 2026
- [People/Entities ] speaker and affiliated organizations context – Juan Andrés Guerrero-Saade (JAGS), OpenAI
Read more: https://www.sentinelone.com/labs/labscon25-replay-keynote-steps-to-an-ecology-of-cyber/