What is AWARE?
AWARE is infrastructure for governing autonomous AI agents. It sits between an agent and the tools that agent calls, and it answers five questions on every call:
- Who is calling? A non-human identity with a credential that is stored only as a hash.
- Is the tool known, and how risky is it?
- What data tier is involved — PUBLIC, INTERNAL, CONFIDENTIAL or RESTRICTED?
- Does a policy govern this agent and this tool?
- What was decided, and can that decision be checked later?
The five layers
| Layer | Responsibility | State |
|---|---|---|
| Identity | Non-human identity, hashed credentials, sessions, revocation | implemented |
| Policy engine | Deny-by-default evaluation of every tool call | implemented |
| T0-T4 | Declared autonomy posture and promotion record | partial |
| Revocation | Immediate credential invalidation and audit | implemented |
| Audit & evidence | Hash chain, independent verifier, exportable windows | implemented |
| Compliance mapping | Seven frameworks, control-to-control mapping | implemented |
| Delegation | Hierarchical authority and cascade | roadmap |
| Execution assurance | Atomic commit, target-side enforcement, verification | roadmap |