AI-Supported Risk Engine for Regulatory Enforcement

Remove Dangerous Non-Compliance From the Market Before It Causes Harm

AI does the assessment — ranking products by risk so your resources go where they matter most — but it doesn't decide anything. It applies your policies, not its own.

02–04Data, fixed before any model runs 05–09AI drafts and verifies automatically, with evidence 10–11Regulator decides — targeting dangerous non-compliance
Probability of non-compliance Consequence if it occurs — these two colors track the two axes throughout
01 · The shape of a decision

Two numbers, plotted, never merged: how likely a product is to be non-compliant, and how dangerous it would be if it is

Every regulated product-type instance an agency oversees ends up as one point on a grid like the one below. Nothing downstream multiplies the two axes together into a single "risk score" — an agency works its queue by scanning the plane, not by reading down a sorted list. The highlighted case is built up step by step across the rest of this page.

Exhibit — one regulator's queue, five real assessed products

The color wash behind each cell is a reading aid, not a stored number — nothing here averages the two axes into a value that gets saved or sorted. These five points follow the same structure and scoring as a real assessment run, with illustrative specifics.

Data

Your rules, fixed first

Your taxonomy, your objectives, your risk-criteria matrix — approved once, up front. No model touches or redefines them; they're simply what gets scored against.

AI

Drafts every score, checked automatically

Two model passes assess likelihood and severity for every product, each producing a factor score or a scenario with the one-line evidence that justifies it. Structural checks, litmus tests, and deterministic recomputation stand between a draft and a stored result — before you ever see it.

Regulator

You decide, always

The assessment is drafted for you, with evidence behind every score. Nothing here acts on your behalf — you're the one who reads the plot, weighs the case, and decides what happens next.

02 · Regulatory universe

Every regulator's world is fixed before a model ever runs

The taxonomy — categories, families, individual product types — is reference data a regulator approves once, up front. No model call touches it. It's simply what the AI layer is scoring against.

Exhibit — a product safety regulator's regulatory universe (general-goods categories shown)
Regulated categoryFamiliesProduct types
Construction materials & building products512
Chemicals & chemical products515
Electrical & electronic products37
Cosmetics & personal care products16
4
Categories shown
40
Product types shown
~140
Full taxonomy, example scale

Cement sits in Mineral construction materials — one of 5 families under Construction materials & building products, one of 12 product types in that category, and one of 40 shown here. A regulator's full taxonomy might run to, for example, 140 product types across a dozen-odd categories; food, beverages, and agricultural inputs would make up the rest and sit outside this page's scope. This case is built up step by step from Section 07 onward.

Plant & Produce Standards Authorityan agricultural / phytosanitary regulator, for comparison
5 categories34 product typesno shared taxonomy with the product safety regulator
A completely unrelated regulated world, run through the same AI-and-code pipeline described below.
03 · Objectives, fixed in advance

What "consequence" means is agreed before the first case

Each dimension of harm comes with four pre-written levels, Low through Very High, approved by the regulator ahead of time. A model drafts scenarios later — it never gets to define what "High" means. The two ladders below are reproduced verbatim from an approved risk-criteria matrix.

Consumer Health & Safety

Very HighFatality, permanent disability, life-changing injury, or widespread exposure to serious hazards.
HighSerious injury requiring hospitalisation or significant medical treatment.
MediumInjury requiring medical treatment but not hospitalisation.
LowMinor injury requiring first aid, or no injury but limited exposure to risk.

Fair Competition

Very HighSignificant market distortion giving non-compliant operators a substantial competitive advantage.
HighSignificant competitive disadvantage imposed on compliant businesses.
MediumModerate competitive advantage gained through non-compliance.
LowLimited competitive advantage with minimal market impact.
The regulator scores four further dimensions the same way: Environmental Impact, Economic Impact, Trade Impact, and Consumer Confidence. Shown in full above for two, to keep this page short rather than exhaustive.
04 · Factor libraries

A fixed list of "why," scored per case

Behind the two axes sit two catalogues: probability factors (why non-compliance might happen) and severity factors (why it would matter). Both are regulator-owned and fixed. What's not fixed is which factors apply to a given product — that's the model's first job, next section. Codes and wording below follow the structure and style of a regulator's own factor lists.

Probability factors — illustrative selection

Consumer Behavior
CB1Consumers willing to compromise on quality in exchange for lower price
CB3High demand / project deadlines encourage shortcuts (construction sector)
Economic Incentives
EI1Significant cost difference between compliant and non-compliant versions
EI3Counterfeiting or false certification (CE, DoP, labels) is economically attractive
Market Structure
MS1Product produced by many small or unknown manufacturers
MS4Product sold mainly in small shops or informal outlets

Severity factors — 6 dimensions, 39 factors (8 shown)

Consumer Health & Safety
S1Product carries high inherent safety expectations — users take fewer of their own precautions
S3There is a compliance requirement that is the primary or only barrier between an inherent product hazard and harm reaching the user — no redundant safety layer exists beyond it
Fair Competition
FC1Compliance cost is a large share of total product cost, so skipping it creates a large price/margin advantage
FC3Compliant firms have made significant sunk capital investment specifically to meet compliance requirements, creating an asymmetric disadvantage
Economic Impact
EC2Product is a critical input to downstream economic activity — remediation/replacement cascades beyond the immediate transaction
EC3Remediation/replacement cost from the non-compliance is disproportionate to the product's own value
Consumer Confidence
CC2Consumers cannot independently verify compliance themselves — trust is entirely delegated to the label/certification this non-compliance undermines
CC3Product is a highly visible, widely recognized consumer staple, so a confirmed incident reaches broad public attention
05 · Two agents, one job each

Consequence gets drafted by two passes, not one

Every assessed instance goes through two model calls before a severity number exists. One covers ground; the other goes deep on what that ground implies.

Agent 1 — breadth

Scores every factor

Checks every severity factor in the library against this product type in general, and may propose factors genuinely missing from the list — a real gap, not a hallucination to be trusted blindly (Section 06 covers how that gets checked).

Ungrounded — reasons from the fixed factor library
Agent 2 — depth

Drafts the scenario

Takes Agent 1's applicable factors as given input and constructs plausible non-compliance scenarios, scored against the regulator's own Low–Very High matrix.

Ungrounded — same reasoning basis as Agent 1

A third, separate call scores probability — and unlike these two, it's grounded in live web search with real citations. See Section 07.

06 · What the model doesn't decide

Four guardrails between a draft and a stored number

"AI-supported" doesn't mean the model's output is taken on faith. Every draft passes through checks that are literal code, not further prompting.

Guardrail 01

The litmus test

A factor only counts if the harm is contingent on the compliance gap itself — not on the product's inherent nature.

A solvent's flammability isn't a factor. A missing child-resistant closure that was supposed to contain it is.
Guardrail 02

Severity anchors

The count of triggered factors in a dimension sets a floor on that dimension's level — thresholds derived from the regulator's own factor-list size, never a fixed table.

A model can't score most of Consumer Health & Safety's factors as applicable and still land the dimension on "Low."
Guardrail 03

Structural validation

Every response is checked against a strict schema — required fields, valid factor codes, and a value/score consistency rule — before it's accepted.

A factor marked "applies" needs a score ≥ 0.5, or the whole call is regenerated from scratch, up to three times.
Guardrail 04

Deterministic recomputation

Aggregation, anchor-floor checks, and every continuous total are recalculated in Python from the model's raw output.

Nothing the model claims about its own total is stored — only what the code recomputes from its factor-level answers.
07 · Scoring probability

Cement: how likely is non-compliance?

Every probability factor is scored 0–1 for this specific product type, each with a one-line evidence note. Unlike Agent 1 and Agent 2, this call is grounded in live web search. This is a real run: the six factors below are a sample of the 22 actually scored.

Drafted by the probability call Real assessment output
0.49
LowMediumHighVery High
Probability index — lands in the Medium band
CB1
Consumers willing to compromise on quality in exchange for lower price
There is strong incentive in the construction market to procure low-cost cement, sometimes leading consumers to accept quality compromises to save on project budgets.
0.70
EI1
Significant cost difference between compliant and non-compliant versions
The cost of producing substandard or adulterated cement is considerably less than maintaining compliance, making non-compliance economically attractive.
0.80
MS1
Product produced by many small or unknown manufacturers
Cement in Malawi is produced and imported by a mix of established and lesser-known suppliers; inspections show many unknown or smaller entities involved.
0.80
SH1
Increase in non-compliance cases compared to previous year
Inspection records indicate numerous non-compliance findings, with a relatively high and possibly increasing number of recent fail/withdrawal outcomes.
0.75
MS6
Product is predominantly sold through online marketplaces
The majority of cement sales in Malawi occur through in-person transactions, not online channels.
0.10
TS6
No recognized standard or regulation applies to this product category at all
Cement is a well-regulated category in Malawi with clear standards in place.
0.05
§

0.49 isn't the average of the six factors above. It's what Guardrail 04 recomputed in Python from all 22 factors in the library — including several, like MS6 and TS6 here, that correctly score near zero because they simply don't apply to cement. Showing only the high scorers would have overstated it. The index never touches consequence — it judges how urgently to look here at all, a separate question from what happens if it goes wrong.

08 · Scoring consequence

Cement: how bad, if non-compliance reaches the market?

Consequence is established twice, independently, and cross-checked — Agent 1's factor pass, then Agent 2's scenario, each held to the guardrails from Section 06. This is a real assessment: the factors and scenario below are reproduced from an actual run.

Severity factor assessment, by dimension
Drafted by Agent 1 — breadth passPassed the litmus test (Guardrail 01)
S1
Consumer Health & Safety — high inherent safety expectations, users take fewer precautions
Cement is assumed to meet reliability and safety standards in construction; users rarely test or question its compliance, relying on standards as a primary protection.
0.80
S3
Consumer Health & Safety — compliance is the primary or only barrier, no redundant safety layer
Compliance with physical and chemical requirements is often the only safeguard — if structural integrity or chemical limits are not met, there is little secondary protection against collapse or toxic exposure.
1.00
FC1
Fair Competition — compliance cost is a large share of total product cost
Quality and compliance testing, as well as materials sourcing, are significant in cement production; skirting standards can provide large market advantages.
0.70
FC3
Fair Competition — compliant firms carry sunk investment, creating an asymmetric disadvantage
Major compliance investments in plant, process control, and quality assurance create disadvantage for compliant producers if others ignore standards.
0.75
EC2
Economic Impact — critical input to downstream activity, remediation cascades beyond the transaction
Cement is foundational in construction; remediation can cause widespread project delays and cost overruns.
1.00
EC3
Economic Impact — remediation cost disproportionate to the product's own value
Replacing or repairing structures due to substandard cement is vastly more expensive than the cement's value.
0.95
Dimension ledger — continuous score (all factors, Guardrail 04) vs. the regulator's ordinal level (worst drafted scenario)
Consumer Health & Safety
0.65 V.High
Fair Competition
0.68 V.High
Economic Impact
0.93 V.High
Scenario S2 · Agent 2 — depth pass, cross-checked against the factors above
Cement is fraudulently diluted with inert fillers or substituted with industrial waste, resulting in weaker-than-stated structural performance. Structures built with this cement are at risk of premature cracking, spalling, or catastrophic failure — potential for building collapse causing fatality or life-changing injury to users and bystanders (S1, S3). Firms cutting composition costs unfairly dominate the market, deeply distorting competition (FC1, FC3), while failure necessitates demolition or costly remedial works with market-wide economic losses affecting the construction sector and downstream users (EC2, EC3).
Drafted by Agent 2 Precedented cites S1, S3, FC1, FC3, EC2, EC3 Fair Competition → Very High
09 · Two readings of severity

Consequence isn't one number even within itself

The regulator's ordinal matrix level reflects the single worst scenario Agent 2 could construct — for Cement, three of the six dimensions reach Very High this way. The continuous consequence index reflects how broadly the severity factors apply overall, across all 39 factors in the library. They diverge, and both get reported rather than reconciled into one.

Ordinal matrix level
Very High
Agent 2's worst-case scenarios — dilution/substitution (S2) and performance failure (S3) both drive Consumer Health & Safety, Economic Impact, and Fair Competition to Very High.
Continuous consequence index
0.72
Agent 1's breadth pass — every one of the 39 applicable severity factors, weighted and normalized against the regulator's full factor library.

Neither number overrides the other. Cement's continuous index (0.72) sits in the "High" band on its own, while its ordinal level is a full step higher at "Very High" — driven by what the worst plausible scenario looks like, not by how many factors apply on average. A regulator needs to see both to know which is true here.

10 · Where the numbers meet

Back to the grid — risk-based, never blended

Probability 0.49. Consequence 0.72. Neither number changes the other, and neither gets multiplied into a single "risk score." Plotted together against four other real assessments from the same regulator's queue, Cement sits high on both axes — exactly the information a blended score would have thrown away.

Exhibit — same queue, Cement highlighted

The color wash behind each cell is a reading aid, not a stored number — nothing here averages the two axes into a value that gets saved or sorted.

Wooden Clothes Pegs sits low on both axes and can wait for a routine cycle. Cement and Insulated Electrical Cable land close together in the upper-right — not because they share a hazard, but because two very different patterns each push both axes up independently: weak testing-and-standards coverage for cement, a fragmented and unregistered manufacturing base for cable.

Not the same as a "most severe scenario" ranking. A separate ranking, computed only within one regulator's results, orders products by how bad their single worst drafted scenario is — useful for picking representative cases when choosing what to test, not for deciding what to check first. Targeting runs off the continuous, factor-derived numbers plotted here instead.

11 · From evidence to test

Every recommended test traces back to a scenario above

The scenario, not a generic checklist, decides what the inspection floor should actually check for. Both tests below are real recommendations from the same Cement assessment.

Analytical testing to verify declared cement composition and detect unauthorized fillers or substitutes (e.g. XRF/compositional assay)
Confirms product composition matches specifications and exposes dilution or adulteration practices.
Addresses Scenario S2 · composition substitution
Very High
Traceability audit — verification of batch, origin, and production records against physical shipment and documentation
Verifies the ability to trace non-compliant product batches and implement targeted recalls.
Addresses Scenario S6 · traceability failure
Medium
Summary

Three actors, not one AI black box

Data

Fixed before any model runs

Every regulator's taxonomy, objectives, and factor library is approved reference data. Nothing here carries over between agencies, and nothing is invented per case.

AI

Drafts, then gets checked

A breadth pass, a depth pass, and a grounded probability call — each producing a score or a scenario with the sentence that justifies it, some backed by live search citations. Structural validation, litmus tests, severity anchors, and deterministic recomputation stand between a draft and a stored number.

Regulator

Decides, always

Probability and consequence stay separate, always, and neither gets acted on automatically. The assessment is drafted for you — you're the one who reads it and decides what happens next.