Observed arrival · 2026-09-20
OSAFIS Maps the Security Boundaries of Intelligent Systems
An open research framework for assessing AI security across nine domains, 23 security properties, and 12 attack mechanisms.
- For
- AI security researchers and assessment engineers
- Worth noticing
- Its nine domains are stable identifiers, not a severity ranking or execution sequence.
Field notes
OSAFIS organizes its proposal around typed relationships such as delegated authority, actions, feedback, and state synchronization. The site distinguishes security properties from intervention mechanisms, test outcomes, evidence facets, and registry decisions, while warning that its architectural examples have not been executed or validated. A 13-document research library includes threat modeling, assessment methods, version identifiers, related frameworks, and worked examples. The current release is labeled 2.0.0-draft.1.
Observed signals
Read the marks
Editorial observations of this landing page, not a rating.
One card from the complete issue