Line
AI Governance
LiveAI Governance makes artificial intelligence governable as an operating system. ENGRAM registers every model, agent and skill with a mandate, a risk boundary and an earned trust level; AI Guardian reviews Fidevo's own AI-assisted work, including this website. Authority comes before autonomy and evidence before trust. Both are in daily internal use, which is why this line is marked live rather than promised.
ENGRAM
ENGRAM is the operating model. It holds five parts that work together.
- Conductor
- Routes work to the capability that holds the mandate for it, and refuses work that has none.
- Skill Registry
- One record per model, agent and skill: owner, mandate, risk boundary, source and timestamp.
- Trust Pyramid L1 to L5
- Every capability starts at L1, sandbox only with synthetic non-sensitive data, and is promoted on evidence.
- Quality Ledger
- Reliability multiplied by usefulness must reach 0.75 or above before a promotion is considered.
- Logos audit pattern
- Each decision keeps the reasoning trail that produced it, so a review can read the argument, not just the output.
AI Guardian
AI Guardian reviews what was actually produced, not what a model was asked to produce. It reads the delivered artefact against the contract it was meant to meet, records the gap where one exists, and attaches owner, source and timestamp to the finding. Fidevo's own AI-assisted work, including this website, passes through it.
Products
- Live
ENGRAM site
The public account of the operating model.
- Live
ENGRAM Trust Portal
Registry, trust levels and quality ledger in daily internal use.
- Early access
AI Guardian
Review of AI-assisted work against what was actually produced.