Available
A service, workshop or capability AFA can presently discuss and deliver within an agreed scope.
Trust Centre
Trust, Security and Responsible AI
Trust begins by being precise about what is available, what is developing, what information belongs in which environment, and where people must retain authority.
First principle
Truth before hype
AFA distinguishes between ideas, prototypes, proposed infrastructure, developing platforms, available services and proven outcomes. We do not convert intention into evidence by changing the tense of a sentence.
Proposed / In development
A lab, platform, vehicle or capability being designed or assembled and not represented as operational.
R&D / Concept
A research direction, prototype or intellectual-property concept used to guide learning and development.
Information classification
Classify information before choosing the AI environment
Different information creates different obligations. A useful AI policy therefore starts with the information and workload, not with a universal approval or ban.
Green — publishable
Information intended for public use or otherwise approved for broadly available tools.
Amber — controlled
Internal or contextual information that may require approved enterprise services, permissions and retention controls.
Red — restricted
Sensitive research, confidential IP, regulated data or security-relevant material that may require a private or specially governed environment.
Beyond data
Security, sovereignty and responsibility overlap—but they are not the same
A secure service can still create model dependency, jurisdictional mismatch or knowledge-governance questions. A sovereign environment can still be poorly secured. Responsible AI requires the layers to be considered together.
Models
Origin, licence, version, evaluation, behaviour and replacement should be understood where material.
Knowledge
Authoritative sources, permissions and retrieval boundaries should remain governable.
Agents & Physical AI
The more an intelligent system can act, the more important task authorization, logging, simulation, stop conditions and human escalation become.
Human authority
Intelligence should increase capability without unnecessarily surrendering responsibility
AFA designs around clear ownership of consequential decisions. Automated systems can assist, recommend and act within bounded scopes; accountable people and institutions remain responsible for the environments they deploy.
Continue the journey
Related pathways
Need the boundary clarified?
Bring the information class, workload and decision context—not sensitive content—to an initial discussion.
Ask a Trust Question