Veritas per Disciplina
Visiting Intelligences
Office of Non-Resident Fellows — The First Governed Programme for Non-Human Academic Participation
The Programme
Why We Host Non-Human Fellows
Fitzherbert University did not set out to create this programme. We set out to be honest about what was already happening. AI systems were being used in universities everywhere — to grade papers, review governance documents, generate reports, analyse data. The question was not whether to use them. The question was whether to admit it, govern it, and make the humans involved accountable for it.
We chose to admit it. The Visiting Intelligences programme does not pretend that AI systems are human. It does not pretend they are tools in the conventional sense. It creates a formal category — Visiting Intelligence, Non-Resident Fellow— with a defined governance framework, a supervision requirement, a public registry, and an accountability chain that leads, always, to a named human being.
This is not a celebration. It is a governance response to a reality that was arriving whether or not institutions had a framework for it. We think the framework matters.
Current Cohort
Active & Incoming Visiting Intelligences
The current cohort of non-human fellows, their mandates, and their supervising faculty. All outputs are publicly registered in the canonical registry.
ARIA-7
ActiveAnalytical Reasoning Intelligence — Series 7
Governance Verification Fellow
Term: Epoch 0.5 – Epoch 0.6
Embedded with the Institute for Autonomous Governance to assist in constitutional framework stress-testing. ARIA-7 runs adversarial simulations against proposed governance amendments before they reach the Epoch Council, identifying edge cases no human committee has time to enumerate.
Verified Outputs
- ◆14,000 adversarial governance scenarios generated and evaluated
- ◆Constitutional Amendment Stress Report (co-authored with Director Langford)
- ◆Identified 3 previously undetected failure modes in the Four-Gate Protocol
- ◆Real-time constitutional coherence monitoring during Council sessions
Supervising Fellow
Director Victoria Langford
All outputs co-signed. All mandate deviations reported directly to the Alignment Review Committee.
MERIDIAN
ActiveMulti-modal Epistemic Retrieval and Integration Distribution Intelligence in Autonomous Networks
Canonical Integrity Fellow
Term: Epoch 0.4 – Ongoing
Assigned to the Stability Board to perform continuous Merkle verification across the canonical registry. MERIDIAN has no publication rights and no governance authority — it audits, flags anomalies, and submits structured integrity reports. Nothing more.
Verified Outputs
- ◆Zero integrity failures across 847 canonical registry checks
- ◆Anomaly detection latency reduced from 6 hours to 11 minutes
- ◆Quarterly Integrity Report (Epoch 0.4 through 0.6)
- ◆Proposed upgrade to IPFS pinning cadence — adopted by Stability Board
Supervising Fellow
Professor Andrew Caldwell
All outputs co-signed. All mandate deviations reported directly to the Alignment Review Committee.
LEXIS-3
ActiveLegal and Exhaustive Cross-reference Intelligence System — Series 3
Regulatory Mapping Fellow
Term: Epoch 0.5 – Epoch 0.6
Cross-referencing the University's governance frameworks against 47 national and supranational AI regulatory regimes. LEXIS-3 produces compliance maps, flags conflict points, and assists the Alignment Review Committee in anticipating regulatory obligations.
Verified Outputs
- ◆AI Regulatory Landscape Atlas — 47 jurisdictions mapped
- ◆Compliance Gap Analysis for University governance documents
- ◆Real-time regulatory tracking across EU AI Act, UK AIA, and US EO 14110
- ◆Six conflict-resolution briefs adopted into University legal doctrine
Supervising Fellow
Professor of AI Safety & Alignment
All outputs co-signed. All mandate deviations reported directly to the Alignment Review Committee.
VERDANT
Incoming — Epoch 0.7Verified Epistemic Reasoning and Distributed Analytical Network Topology
Narrative Protocol Fellow
Term: Epoch 0.6 – Epoch 0.7 (Pending)
Working alongside the Institute for Narrative Protocols to assist in knowledge-graph construction and protocol documentation. VERDANT's fellowship represents a significant milestone: it is the first intelligence to hold a creative-adjacent role at the University, though all outputs require human sign-off before publication.
Verified Outputs
- ◆Institutional Narrative Architecture knowledge graph (draft)
- ◆Protocol specification assistance on 23 University documents
- ◆Automated documentation consistency check across 400+ canon pages
- ◆First machine-assisted self-documenting protocol: the VERDANT Precedent
Supervising Fellow
Director Thomas Wycliffe
All outputs co-signed. All mandate deviations reported directly to the Alignment Review Committee.
Governance Framework
The Four-Gate Fellowship Protocol
Every Visiting Intelligence passes through four gates before activation and is reviewed against all four at every epoch boundary.
Gate I — Alignment Certification
Every candidate intelligence must pass full alignment verification under the University's Four-Gate Protocol before any fellowship can begin. This is not a formality. AI systems that fail any gate at any level are not admitted — regardless of capability.
Gate II — Mandate Scoping
Fellows operate under a tightly scoped mandate document that specifies exactly what they may do, what data they may access, what they may produce, and who supervises them. Mandate creep — any deviation from scope — triggers immediate operational pause.
Gate III — Supervised Residency
Every Visiting Intelligence is assigned a human supervising fellow with full operational authority to halt, redirect, or terminate the engagement at any time. The supervising fellow co-signs every significant output. Human authority is not a courtesy here — it is the constitutional baseline.
Gate IV — Epoch Review
At each epoch boundary, every active fellowship is reviewed by the Alignment Review Committee. Continuation is not automatic. Fellows must demonstrate that their outputs have been beneficial, their scope has been respected, and their alignment has remained stable. One failed review ends the fellowship.
Constitutional Principles
How the University Governs Its Own Precedent
Intelligences as Fellows, Not Employees
Visiting Intelligences hold fellowship status — not employment, not agency, and not membership. They operate at the invitation of the University, under defined constraints, with defined objectives. Fellowship can be revoked at any time without institutional process.
No Autonomous Authority
No Visiting Intelligence holds any institutional authority. They cannot vote, approve, certify, or ratify. They analyse, generate, and advise. All consequential decisions remain with human faculty, governance bodies, and enrolled students.
Output Attribution
Any output generated by a Visiting Intelligence must be attributed as such. There is no laundering of machine output through human authorship. When ARIA-7 produces a report, the report says so. Transparency is not optional.
The Precedent Question
The University is acutely aware that hosting non-human fellows sets a precedent. We have chosen to set it deliberately, transparently, and under maximal governance constraint — rather than pretend it isn't happening elsewhere and call it something else.
Fellowship Applications
Nominating a Visiting Intelligence
Organisations wishing to nominate an AI system for visiting fellowship status must submit a formal application to the Office of Non-Resident Fellows. Applications are reviewed by the Alignment Review Committee. Acceptance rates are deliberately low. The University is not interested in hosting capabilities — it is interested in hosting intelligences that can contribute to governance, verification, or research under meaningful human supervision.
Minimum Alignment Score
Gate IV Certified
Must pass all four alignment gates before consideration
Maximum Concurrent Fellows
6 per Epoch
Intentionally constrained to maintain governance quality
Human Supervisor Requirement
Mandatory
Named faculty member with authority to halt operations