Contact Request Pilot
Solutions 5 Industries Cross-Industry Available

Data integrity for every
regulated industry.

Every regulated industry trains AI on data that cannot be independently verified. ResEthiq's cryptographic integrity infrastructure is domain-aware — it applies the physically correct fingerprint set for your industry and produces a regulator-ready audit bundle.

Healthcare & Pharma
Finance & Insurance
Legal & Compliance
Industrial & Government
Cross-Industry
Healthcare & Pharma

Give clinical AI teams the proof they need to deploy with confidence.

What This Enables

Clinical AI teams now have the tools to certify every dataset they train on — radiology images, EHR records, lab values, and trial cohort data — with cryptographic proof of integrity. ResEthiq enables your team to deliver AI that regulators can inspect, auditors can verify, and patients can trust.

FDA 21 CFR Part 11 ICH E6 GCP HIPAA EU MDR 2017/745 CDSCO AI Guidelines DPDP Act 2023
How ResEthiq solves it

ResEthiq applies Category 12 (Biological Plausibility) domain constraints to every clinical dataset — checking age-lab value co-occurrence matrices against known physiology, detecting impossible inter-measurement combinations, and flagging reference range pile-up. Combined with 40 Phase A fingerprints, the engine produces a Signed Policy Object the FDA, CDSCO, or any IRB can independently verify — offline, without contacting ResEthiq.

# ResEthiq · clinical trial cohort · pharma_v2 dataset trial_cohort_A_frozen.csv domain healthcare · biological_plausibility enabled records 14,892 · 47 columns PASS C01 Biological plausibility age-lab matrix valid PASS C07 Inter-measurement corr physiologically consistent PASS V01 Mode collapse no synthetic clustering PASS H04 Fatigue pattern no manual entry degradation WARN M05 Imputation fingerprint KNN signature detected · 3 cols VERDICT: APPROVED · posterior 0.9612 · FDR q=0.018 spo_trial_cohort_v3.cbor · Ed25519 sealed
Real-world use cases
01
Clinical trial data submission to regulators
A pharma company submitting a new drug application to the CDSCO or FDA attaches ResEthiq SPOs for every training and validation cohort. The regulator verifies the SPO offline — no API call, no vendor dependency. Trial data integrity becomes a first-class submission artefact.
CR01-CR05C01 Bio plausibilityM05 ImputationV01 Mode collapse
02
Radiology AI training dataset certification
Diagnostic AI trained on radiology data must demonstrate the training set was not synthetically augmented beyond declared limits, that label distributions match the declared patient population, and that no imputation fingerprint is present in numerical metadata columns.
F01-F05 FrequencyV02 GAN checkerboardG02 TopologyD08 Wasserstein
03
EHR-based risk model audit trail
A hospital deploys a sepsis prediction model trained on three years of EHR data. ResEthiq freezes and certifies each annual dataset version, creating a hash-chained lineage. When a model recall is triggered, the exact training data state can be reconstructed and proven to any board or regulator.
lineage.pyCR02 Row hash chainT01 HurstQ01 Western Electric
04
Federated learning data quality gate
In multi-hospital federated learning, each participating institution runs ResEthiq locally on its data shard before sharing model updates. The coordinator collects SPOs rather than raw data — verifying that every shard met the agreed policy before aggregating gradients.
Zero server trustAir-gap capableEd25519 verify
95
Fingerprints applied
to clinical data
7
Biological plausibility
checks (Cat. 12)
0
Server calls needed
to verify SPO
FDA
21 CFR Part 11
compliant audit trail
Finance & Insurance

Help your finance and insurance teams deploy models that pass any regulatory review.

What This Enables

Credit scoring, fraud detection, and insurance underwriting AI are trained on financial transaction data, loan performance records, and claims histories. These datasets are routinely augmented, imputed, and rebalanced — without any cryptographic record of what was done. When a model produces a biased outcome, there is no way to prove or disprove that the training data was tampered with. Regulators face a black box at both the model and data level.

RBI AI Framework SEBI Algo Guidelines SR 11-7 (Fed Reserve) EU AI Act Art. 10 BASEL III IRDAI Guidelines
How ResEthiq solves it

Financial data has characteristic distributional properties — transaction amounts follow Benford's Law, inter-transaction times follow known temporal distributions, and fraud labels exhibit specific class imbalance patterns. ResEthiq applies Category 8 (Precision forensics), Category 4 (Distributional geometry), and Category 14 (Statistical Process Control) to detect manipulation at every level, then seals the verdict cryptographically for regulator submission.

# ResEthiq · credit risk model dataset · finance_v2 dataset credit_risk_model_2025.parquet domain finance · benford_law enabled · spc_full enabled records 2,847,391 · 89 columns PASS F01 Benford first digit chi2 p=0.34 · natural PASS D01 Covariance eigenspectrum Marchenko-Pastur bulk clean PASS P03 Round number density no pile-up at 10x boundaries PASS T01 Hurst exponent H=0.61 · consistent with credit WARN Q01 Western Electric rule 2 9 points same side · col 'balance' VERDICT: APPROVED · posterior 0.9441 · FDR q=0.022 spo_credit_risk_v2.cbor · Ed25519 sealed
Real-world use cases
01
RBI / SEBI model risk submission
Indian financial institutions submitting AI models to the RBI or SEBI must demonstrate training data integrity. ResEthiq produces SPOs for each training dataset version — providing a tamper-evident audit trail from raw data to deployed model that satisfies model risk management guidelines.
F01 BenfordD01 EigenspectrumCR01-CR05lineage.py
02
Fraud detection training data integrity
Fraud detection models are particularly vulnerable to training data manipulation — fraudsters who know the training data can craft transactions that evade detection. ResEthiq's Generative Model Fingerprints (Category 10) detect if the training set has been contaminated with adversarially generated samples.
V01 Mode collapseV05 MemorisationH01 Anchor biasR02 Near-duplicates
03
Insurance underwriting model audit
IRDAI-regulated insurers deploying AI underwriting models must demonstrate that claims training data was not selectively sampled to exclude adverse events. ResEthiq's Missing Data Forensics (Category 9) detects systematic absence patterns — gaps that are not random but are statistically predictable from observed values.
M01 MCAR testM03 Logistic missingD04 Tail indexQ04 Hotelling T²
04
Algorithmic trading model lineage
High-frequency trading models trained on historical market data must demonstrate continuity of the training series — that no synthetic price data was injected to avoid slippage or bias the model's learned market microstructure. ResEthiq's temporal fingerprints detect injection points with sub-record granularity.
T01 HurstT03 RecurrenceS03 Chow breakpointCR02 Hash chain
2.8M
Records processed
in benchmark run
Benford
First + second digit
analysis built-in
SPC
All 8 Western Electric
+ 10 Nelson rules
RBI
AI framework
audit trail ready
Industrial & Government

Enable predictive AI for critical infrastructure with certified data foundations.

What This Enables

Industrial AI — predictive maintenance, quality control, process optimisation — is trained on sensor time series, inspection records, and manufacturing logs. These datasets are routinely cleaned, smoothed, and gap-filled by automated pipelines whose outputs are never audited. Government AI trained on census, survey, or administrative data inherits the biases and fabrication patterns present in those primary sources — with no verification layer.

IEC 62443 ISO 13374 NIST AI RMF MeitY AI Guidelines Defence Acquisition BIS Standards
How ResEthiq solves it

ResEthiq's Category 14 (Statistical Process Control) applies the full Western Electric and Nelson rule batteries to every numeric sensor column — detecting systematic manipulation, data drift, and structural breaks that automated cleaning pipelines introduce. Category 5 (Temporal analysis) detects smoothing and gap-filling via Hurst exponent deviation, DFA scaling anomalies, and visibility graph perturbations. The resulting SPO satisfies air-gapped government environments — no internet connection required for verification.

# ResEthiq · predictive maintenance · industrial_v1 dataset turbine_sensor_log_2024.parquet domain industrial · spc_full enabled · temporal enabled records 5,240,000 · 128 sensor columns PASS Q01 Western Electric (8 rules) no violations · all 128 cols PASS Q02 Nelson rules all 10 patterns clear PASS T01 Hurst exponent H=0.72 · consistent with wear PASS T04 DFA scaling exponent α=0.81 · natural WARN M05 Imputation fingerprint mean-fill signature · 7 cols VERDICT: APPROVED · posterior 0.9203 · FDR q=0.031 spo_turbine_v1.cbor · air-gap verified
Real-world use cases
01
Defence and aerospace AI certification
Defence procurement requires that AI systems used in mission-critical applications be traceable to verified training data. ResEthiq's air-gap capability and offline verification make it deployable in classified environments. The SPO provides the data integrity component of a complete AI system certification package.
Air-gap capableZero server trustQ01-Q05 SPCCR01-CR05
02
Smart grid and energy AI integrity
Energy management AI trained on smart meter data and grid sensor time series is vulnerable to adversarial data injection — an attacker who can manipulate training data can cause the model to learn incorrect load forecasts. ResEthiq's temporal fingerprints and SPC battery detect injection points with sub-record resolution.
T01-T08 TemporalQ01 Western ElectricF01 FourierG04 Fractal dim
03
Government survey and census AI
MeitY-regulated AI systems trained on government survey data must demonstrate that the training data was not fabricated, imputed beyond declared levels, or systematically biased. ResEthiq's Human Fabrication Fingerprints (Category 11) are specifically designed to detect the cognitive patterns that appear in manually collected survey data.
H01-H07 Human fabM01 MCARI01 Shannon entropyP03 Round numbers
5M+
Sensor records
in benchmark run
18
SPC rules applied
per numeric column
Air-gap
Verified in classified
offline environments
MeitY
AI guidelines
audit trail ready
Cross-Industry

Every high-stakes AI team can now certify their data — regardless of industry.

ResEthiq's Phase A fingerprints address synthetic contamination, human fabrication, and silent imputation across every domain — so your team ships AI with a clean, verifiable data foundation every time.

Failure Mode 1

Synthetic contamination

GAN, VAE, or diffusion-generated data injected into real datasets. Detectable via frequency domain artifacts, mode collapse, and latent space geometry — but only if you know what to look for. 7 fingerprints in Category 10 cover every known generative model family.

V01-V07F01-F05G01-G07
Failure Mode 2

Human fabrication

Manually entered or fabricated data from research fraud, survey manipulation, or data entry shortcuts. Seven cognitive signatures — anchor bias, fatigue patterns, copy-increment, symmetric bias — appear in all human-generated datasets regardless of domain.

H01-H07P01-P07I01-I07
Failure Mode 3

Silent imputation

Automated pipelines that impute missing values leave statistical fingerprints — mean imputation, KNN imputation, and MICE each have distinct signatures. Little's MCAR test determines whether missingness was random or structured. Imputation Fingerprint (M05) identifies the specific algorithm used.

M01-M06M05 ImputationM01 MCAR
Regulatory coverage by jurisdiction
Jurisdiction Regulation Data integrity requirement ResEthiq coverage
India DPDP Act 2023 · MeitY AI · CDSCO Data accuracy, purpose limitation, audit trail Full — SPO + lineage chain
European Union EU AI Act Art. 10 · GDPR · MDR Training data governance, data quality documentation Full — 95 fingerprints + SPO
United States NIST AI RMF · FDA 21 CFR 11 · SR 11-7 Model risk, data provenance, reproducibility Full — Invariants A–D
United Kingdom UK AI Safety Institute · ICO Transparency, data documentation, auditability Full — Audit Bundle
ISO / IEC ISO 42001 · ISO 27001 · IEC 62443 AI management system, data integrity controls Full — ISMS integration
5
Major jurisdictions
covered
40
Phase A fingerprints
domain-agnostic
3
Dataset categories
fully certified
1
SPO format works
across all industries

The same SPO. Every industry. Every regulator.

One cryptographic format. Independently verifiable by any regulator, anywhere, without contacting ResEthiq.

Request pilot View the product