Enregistré dans:
| Auteur principal: | |
|---|---|
| Format: | Recurso digital |
| Langue: | anglais |
| Publié: |
Zenodo
2026
|
| Accès en ligne: | https://doi.org/10.5281/zenodo.20370944 |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
Table des matières:
- <p>Google's recent disclosure of an AI-assisted zero-day exploit is being framed as the moment AI crossed the threshold into autonomous offensive capability. This paper argues that framing is incorrect. The real milestone is not that AI discovered a zero-day — offensive AI capability at this level has likely existed for some time. The milestone is that our defensive instrumentation finally produced enough signal to attribute one. This is a defensive visibility event, not an offensive capability event. The paper examines what that distinction means for security architecture, DFIR methodology, and the epistemic requirements of detection in an era of machine-scale adversarial behavior.</p>