Description
RegenData is a constraint enforcement and audit layer for AI systems operating in water and physical infrastructure domains.
AI systems operating in physical infrastructure can produce confident, fluent outputs that violate the physical constraints of the system they are advising. Existing tools, sensors, control systems, document retrieval, digital twins, improve what AI knows. They do not govern what AI is permitted to conclude from that information.
RegenData addresses that gap through three components. A domain canon encodes the physical, chemical, and biological mechanisms governing a specific infrastructure domain as a structured rule set. An execution boundary evaluates every AI output against the active canon before it reaches an operator or system, and either permits it or blocks it and routes it to human review. A provenance store seals an immutable, tamper-evident record of every gate decision, capturing the input, canon version, constraint rules evaluated, and outcome.
The first deployed canon covers onsite wastewater treatment systems. Pilot validation demonstrated a 76% reduction in unsafe AI recommendations under canon-constrained versus unconstrained conditions using the same model and the same prompts. The constraint layer was the only variable.
RegenData is currently seeking co-development partners in water and wastewater domains to build and deploy canons for adjacent physical infrastructure verticals. The EU AI Act classifies water and wastewater AI as high-risk systems. Enforcement begins August 2026.
The execution boundary and provenance store are currently in active development. Co-development pilot engagements are open now
Target audience
Owner of the product
Sign in to access this information.Actors, their roles and interactions
Water utilities and wastewater operators: deploy RegenData as the constraint and audit layer governing AI-assisted operational decisions before they reach control systems or field operators.
Infrastructure AI companies and platform providers: integrate the RegenData execution boundary as a post-inference constraint layer within existing AI deployments, digital twins, and decision support platforms.
Engineering firms and system integrators: embed the canon architecture within AI-assisted design, assessment, and compliance workflows.
Regulators and compliance bodies: use the RegenData provenance store as a tamper-evident audit trail demonstrating that AI outputs were governed against a versioned domain canon at the moment of decision.
Insurers and risk underwriters: use the atomic decision records to establish a clear, reconstructable liability chain for AI-mediated infrastructure decisions.
Unique selling points
- Deterministic constraint layer: AI outputs are permitted or blocked against domain rules, not probabilistically weighted
- Mechanism-anchored: constraint rules are grounded in the physical, chemical, and biological reality of the domain, not content filters
- Atomic provenance records: every gate decision generates an immutable, tamper-evident record capturing input, canon version, constraint rules evaluated, and outcome
- Works across existing stacks: sits between AI output and real-world action regardless of the underlying model, data source, or platform
- Validated result: 76% reduction in unsafe AI recommendations in pilot study, same model, same prompts, constraint layer as the only variable
- Regulator and insurer ready: audit trail satisfies EU AI Act high-risk AI documentation requirements
- No retraining required: deploys at inference, not at training
Technical requirements
RegenData deploys as a post-inference middleware service via API. Integration requires the ability to route AI outputs through the execution boundary before delivery to operators or systems. Canon deployment is a configuration load. No model retraining or modification required.
Software data
| 0.1 | |
| 2026 | |
|
|
| Commercial |
Publications
Box, C. (2025). Mechanism-Anchored Constraint Architecture (MACA): A Framework for Governing AI Outputs in Physical Infrastructure Domains. RegenData Technical White Paper. https://github.com/Colin-io/RegenData-v2-whitepaper/blob/main/RegenData_V2_Technical_Paper.pdf