RATE AI
EN·DE·FR·ES
Reports
Human Intelligence for AI · Independent Monitoring

Know when human judgement needs renewed attention

RATE AI connects verified changes in systems, providers, regulation, enforcement, incidents, evidence and dependencies to the exact oversight perimeter, decision trace and resilience question they can affect.

Verified change is linked to the exact systems, deployments, jurisdictions, evidence and accountable owners it affects, then assessed for materiality and the appropriate review trigger.

MATERIAL CHANGE ALERT

A change matters when it changes the human judgement required

A material alert connects the event to affected systems, human oversight perimeters, decision traces, dependencies, owners and review dates.

What changedProvider version update
Affected perimeter2 systems · 5 deployments · 3 decision traces
Evidence affectedValidation record · supplier evidence
Decision conditionC2 · version-specific validation
MaterialityREVIEW
ActionConfirm validation scope before the next rollout

Observation starts before the client mandate

RATE AI is built on linked collections that support longitudinal observation across systems, evidence and time. The data infrastructure establishes the denominator, preserves provenance and creates a common evidence layer from which systems can be sampled, rated, compared and monitored over time.

Flagship infrastructure

The RATE AI Census

The Census continuously identifies and delimits documented AI systems from public evidence, keeping system identity, provider, deployment context, jurisdiction, source provenance and verification status connected. Its value lies in the observation frame from which RATE AI can validly select, compare, rate and revisit real systems.

Discovery at scaleLarge-system enumeration before rating or sampling.
Evidence linkedRecords retain source provenance and verification status.
VersionableFrozen snapshots make changes measurable across rating generations.

The live Census continues to collect after every freeze. Each rating is calculated from an immutable, versioned analytical snapshot tied to a defined evidence cut-off.

Legal & regulatory

Global Legal Library

Legislation, official guidance, regulator decisions, enforcement, jurisprudence and implementation evidence mapped to jurisdictions and systems.

Ratings & evidence

Rating Library

System-level indicator decisions, evidence states, provenance, cut-off dates, rating history and comparable analytical records.

Human evidence

Humans RATE AI

Reported experience, access conditions, agency, friction and observed effects kept alongside technical, legal and institutional evidence.

Contextual data

Economic & sector data

Official statistics and contextual indicators used to understand scale, strategic domain, institutional capacity and the environment in which deployment occurs.

Dynamic signals

Incidents · enforcement · dependencies

Operational events, recalls, investigations, provider/model changes and supply-chain dependencies that can alter a previously observed position.

The operating field around AI

A material change can begin inside the system, outside the organization, or in the legal and institutional environment around it. RATE AI follows the signals that can change exposure, defensibility or the decision window.

Deployment

System & use change

New deployment, scale, geography, use case, user reach, decision role, configuration or operating context.

Regulation

Rules & implementation

Legislation in force, application dates, delegated or implementing measures, national rules, official guidance and supervisory expectations.

Enforcement

Cases & jurisprudence

Investigations, decisions, sanctions, court judgments, regulator statements and emerging enforcement patterns.

Incidents

Failure & adverse events

Incidents, recalls, corrective actions, safety notices, documented harms, withdrawals and remediation signals.

Technology

Model & version change

Model releases, material updates, capability shifts, performance or safety disclosures and change-control events.

Dependencies

Provider & supply chain

Foundation-model, cloud, dataset, API, component and specialist dependencies that can concentrate or transmit exposure.

Assurance

Standards & evidence

Changes in assurance expectations, testing practice, certification, risk-management standards and evidence requirements.

Human consequence

Reach & reversibility

Signals that change how many people are exposed, how consequential the decision becomes, or how difficult the outcome is to reverse.

How intelligence is qualified

Material change becomes actionable human intelligence

DetectCapture a new fact, rule, event, release, decision or disclosure.
VerifyConfirm source quality, date, jurisdiction, affected unit and evidence state.
ConnectMap the signal to systems, entities, portfolios, industries, sectors or jurisdictions.
Read materialityAssess consequence, reach, timing, dependency and whether control or defensibility changes.
TriggerIssue an alert, open Decision Traceability, renew the oversight perimeter, initiate rerating, or trigger resilience action.

A signal is promoted when verified evidence shows that it can materially change the decision environment.

Method behind the intelligence

Verified evidence. Decision relevance. Action

The intelligence layer and the ratings use the same evidence discipline. Sources are attributed, the exact evaluated unit is defined, evidence is qualified before scoring, missing evidence remains visible, and every analytical generation is tied to a methodology version and evidence cut-off.

CollectCapture source provenance, system identity, deployment and contextual facts.
VerifyCheck source route, applicability and whether the evidence belongs to the exact evaluated unit.
QualifySeparate qualifying evidence, subthreshold evidence, missing criteria and insufficient evidence.
MeasureApply the frozen REI / ADI indicator rules with explicit evidence states and transparent treatment of missing data.
FreezeRecord schema, methodology version, evidence cut-off, snapshot identity and integrity hashes.
ReviewUse HITL, data-integrity checks, challenge logs and independent governance before publication.
Why this matters: change in a RATE AI output must be distinguishable from three different causes: the system changed, the observed universe changed, or the methodology changed. The data and methodology architecture separates these causes so each observed difference can be attributed correctly.

The same observation layer supports different decision lenses

Public Intelligence explains broad movement in the landscape. Commissioned Intelligence narrows the same observation system to the client's actual systems, jurisdictions, dependencies and decision thresholds.

Level
Public intelligence
Commissioned intelligence adds
Global Outlook

Global movement, major regulatory shifts, systemic providers and cross-border patterns.

Custom universe, exposure pathways, dependency concentration, scenarios and implications for the client's footprint.

Jurisdiction Intelligence

Legislation, implementation, enforcement posture, institutional timing and comparative movement.

Client-specific jurisdiction clocks, affected-system mapping, subnational divergence, enforcement scenarios and decision deadlines.

Industry Intelligence

Industry-wide exposure, provider concentration, incidents, regulation and comparative signals.

Peer set, company/value-chain decomposition, material outliers, strategic scenarios and competitive exposure.

Sector Intelligence

Sector-level movement, implementation conditions and emerging consequence patterns across public and private deployment where relevant.

System/deployer decomposition, human-reach analysis, supervisory priorities, evidence gaps and action triggers.

System / Portfolio Watch

Entity-level detail is generally reserved for commissioned intelligence.

Named systems, models, providers, dependencies, incidents, version changes, regulatory impact and event-driven rerating triggers.

Commissioned Intelligence creates a narrower, deeper and continuously updated decision view around the client's own exposure while public rating authority remains unchanged.

Current obligations and emerging requirements are tracked separately

RATE AI separates legal applicability from supervisory interpretation, assurance standards and forward-looking governance signals. That distinction prevents both underreaction and premature over-compliance.

CURRENT / BINDING

Applicable obligation

Rules that are legally or contractually applicable to the assessed role, system, use, jurisdiction and date.

Examples: EU AI Act where applicable, GDPR, DSA, MDR/IVDR, sector and national rules.
SUPERVISORY

Implementation & enforcement

Official guidance, regulator practice, enforcement decisions and implementation expectations that shape how rules operate in reality.

Competent-authority guidance, enforcement posture, court decisions, procurement and supervisory expectations.
ASSURANCE

Standards & evidence practice

Standards that may be voluntary but increasingly determine whether an organization can demonstrate credible governance and assurance.

Examples: ISO/IEC 42001, ISO/IEC 23894 and relevant sector assurance standards.
ANTICIPATORY

Emerging governance signal

Frameworks, revisions, consultations and soft-law signals that may foreshadow future due-diligence or governance expectations.

Examples: NIST AI RMF, OECD, UNESCO, Council of Europe implementation and emerging national initiatives.

RATE AI identifies each framework by legal status, applicability and effective date for the specific mandate.

From verified change to executive action

Intelligence is translated into bounded decision states so the organization knows what deserves attention now.

Act now

Material position change

An incident, control gap, regulatory event or exposure shift requires a concrete decision or intervention.

Prepare

Foreseeable change

The position is defensible today, but an implementation date, system change or emerging obligation requires readiness.

Monitor

Signal under watch

Current evidence supports continued surveillance; specified developments can move the position.

Maintain

Current position

The reviewed evidence supports maintaining the current position. The rationale and evidence cut-off remain recorded.

Human judgement stays connected to evidence, dependencies and change after deployment

Decision TraceabilityOpen a decision-traceability, evidence-readiness or defensibility review when the signal affects a consequential decision.
ReratingRecalculate the independent position when new evidence materially changes exposure or demonstrated control.
Resilience & Operational IndependenceMap the affected dependency, including material Shadow AI exposure, test alternatives and coordinate specialist work when continuity or substitution requires action.
Independence firewall. Continuous Intelligence may trigger renewed oversight, resilience work or rerating. Commercial execution remains separately governed from rating authority. Any new rating arises from independently admitted updated evidence.

Public intelligence shows the pattern. Client intelligence names the perimeter

Public releases show the patterns RATE AI is prepared to put into the market. Confidential mandates go further: entity-specific standing, peer movement, evidence gaps, regulatory pressure and material-change signals prepared for boards, supervisors, investors and legal teams.

Global Outlook

What is changing across AI deployment

Global shifts in deployment, regulation, enforcement, concentration and consequence exposure.

Jurisdiction

Regulatory clocks and implementation divergence

Where national, supranational and subnational timelines are moving at different speeds.

Industry

Emerging exposure and dependency patterns

How model providers, value chains, incidents and regulation are reshaping industry positions.

Enforcement

Cases that can change the decision environment

Material enforcement and jurisprudence interpreted against the systems and uses they can affect.

Public intelligence releases are dated, archived and connected to the relevant rating, evidence cut-off and methodology version.

PositionWhat can be proved today and where the entity sits relative to the field.
MovementWhat changed in the system, evidence, peer group or enforcement environment.
PressureWhat is likely to become material before the next formal review.
Explore The Signal →Discuss a confidential briefing →