Background & research
Confidential computing and AI trust boundaries
How confidential execution and attestation fit into a broader AI assurance design.
Read noteTechnical notes
Notes
Technical notes on trustworthy AI systems, runtime boundaries, memory protection, and infrastructure.
Research notes and design questions from AGIACC. Product directions are identified as planned; third-party technologies are referenced as context. Explore the current development agenda in Research directions.
Background & research
Designing boundaries for tool-using agents
A technical note on containment, authority, and recovery in agent runtimes.
Read noteBackground & research
From runtime security to trustworthiness-native AI
A problem-led view of protection across hardware, systems software, and application frameworks.
Read note