Why it matters
Without this system, foxing and mould issues can advance unchecked, causing irreversible staining, failed treatments, missed loan deadlines, disputed insurance claims, and eroded client trust.
From design to safety
Design
AI-guided treatment planning and policy
The AI compiles historical condition reports, high-resolution baseline scans, literature reviews, and ethical constraints to propose methods, materials, reversibility strategies, and documentation templates before any intervention.
- foxing detector
- tide line monitor
- paper fibres analyzer
Build
System validation and instrument calibration
During build, calibrated imaging, spectroscopic sensors, moisture meters, and data pipelines are verified against reference datasets; calibration certificates, QA run logs, and traceable metadata are archived.
- mould stain mapper
- washing tests controller
- foxing detector
Operation
Active monitoring, imaging, and decision support
The AI analyzes daily high-res scans, colorimetry, humidity data, and surface readings to classify foxing progression, mould extent, tide line shifts, and fibre degradation, updating treatment plans and recording actions.
- foxing detector
- mould stain mapper
- tide line monitor
- paper fibres analyzer
Maintenance
Data governance, stability checks, and archival
Ongoing calibration of imaging and sensors, periodic verification against reference folders, and continual documentation of washing tests and outcomes; all updates are versioned in the project archive.
- washing tests controller
- treatment records integrator
Safety
Monitoring, safety checks, and ethical compliance
The system enforces COSHH-inspired checks, records reversibility and ethical considerations, flags deviations for qualified sign-off, and maintains a chain of custody for materials and records.
- treatment records integrator
- washing tests controller
- paper fibres analyzer
Part by part
- 1
foxing detector
- Watch for
- intrinsic orange-brown spots indicating foxing progression and edge spread toward text blocks with potential pigment interaction
- AI action
- detects early foxing signs, correlates with historic paper type, triggers high-res imaging, and generates a temporary treatment proposal flag
- 2
mould stain mapper
- Watch for
- sporadic circular or irregular staining indicating mould growth and surface bloom on varying paper weights
- AI action
- classifies mould extent, maps infected area, estimates moisture history, and suggests humidification plan with containment actions
- 3
tide line monitor
- Watch for
- tidal staining along margins due to previous damp events, increasing risk of ink bleed or fibre failure
- AI action
- maps tide line geometry, tracks movement over time, and suggests stabilization plan with moisture control steps
- 4
paper fibres analyzer
- Watch for
- edge fiber fraying, surface pitting, and changes in lignin content affecting strength
- AI action
- estimates paper integrity index, guides poultice or washing decisions, and records fibre condition data to the treatment record
- 5
washing tests controller
- Watch for
- solvent interaction, swelling, colour bleed during test wash, indicating cross-reaction with inks or adhesives
- AI action
- predicts compatibility, records results to treatment records, and flags potential corrective actions before bench work
- 6
treatment records integrator
- Watch for
- incomplete documentation, missing reversibility notes, or absent provenance traces
- AI action
- enforces template completion, cross-checks against standards, and suggests missing fields for audit readiness
Worked example
Catching a foxing bloom before a loan
Illustrative scenario with sample data.
An unseen foxing bloom begins to spread on a mid-19th century paper folio scheduled for loan; AI flags early signs, prompts targeted imaging, and generates reversibility notes and a treatment proposal.
Signal in
coatingThickness: 0.12, surfaceRBC: 1.4, foxingScore: 2, scanDate: 2026-05-04, conditionReportID: CR-0771, paperType: handmade
System out
alert: foxing onset detected; action: initiate high-res mapping, prepare treatment proposal; reportID: TP-ON-0771; recommendedMaterials: calcium carbonate poultice, deionized water rinse, reversible starch paste
Human sign-off: Head of Conservation validates the treatment proposal, approves the imaging plan, and signs the revised condition report before bench work begins.
How to roll it out
Phase 1
Weeks 1 to 4
Baseline data capture, sensor calibration, and template alignment with condition reports; establish version control and archive structure.
Phase 2
Weeks 5 to 12
Pilot integration with treatment proposals, imaging workflows, and print-ready records templates; end-to-end provenance chain is activated across cases.
Phase 3
Weeks 13 onward
Full deployment across projects with ongoing QA, archival integrity checks, and routine audits for ethics and reversibility.
Where people stay in charge
- Decisions about treatment necessity, reversibility, and material choice must stay with a qualified conservator and be documented.
- All data handling and privacy decisions must be signed off by the studio’s data protection lead and conservator.
- AI-generated recommendations cannot replace professional judgment on hazardous materials or ethical considerations; only accredited conservators sign off.
Questions owners and crews ask
What is AI foxing detection for works on paper?
AI foxing detection analyzes surface patterns, pigment interactions, historical condition data, and solvent history to flag early foxing signs, guide imaging, and inform treatment proposals with traceable records.
How does mould stain detection work in papers?
Mould stain detection models classify stain morphology, quantify area, relate it to moisture history, and suggest stabilization and cleaning strategies while recording results in the treatment records.
What data are required to start AI monitoring?
Baseline condition reports, high-resolution scans, moisture readings, humidity histories, and material specifications feed the model to generate proposals and tracking records.
How is safety handled with AI in conservation?
Qualified conservators approve all actions, COSHH-informed checks are logged, reversibility remains the guiding principle, and the chain of custody for materials and data is maintained.