Every public claim we make,
versioned and sourced.

Mergetic is designed as an independent runtime AI governance layer. AI decisions routed through an agreed Mergetic deployment pass through a split-authority control point before external effect: evaluation and routing are structurally separate from the governance authority that applies the active RulePack and can issue a binding veto.

Applies to AI decisions routed through the agreed deployment boundary; it is not a claim about AI activity outside that boundary.

The public disposition vocabulary is PERMIT, REGEN and BLOCK. PERMIT releases the candidate action, REGEN requires a constrained replacement, and BLOCK prevents the candidate action from taking effect. A configured workflow can route a blocked action to accountable human review.

A held-for-review effect is recoverable; a hard block is not represented as a release.

Illustrative Mergetic decision records use SHA-256-linked integrity events to demonstrate an append-only, tamper-evident record produced by execution rather than reconstructed afterwards.

Public demonstrations are illustrative and are not production certificates.

Mergetic is designed to be vendor-neutral and to support multiple model and cloud choices without requiring a model swap.

Each model, cloud and integration pattern remains subject to technical validation during scoping.

Mergetic is designed to fail closed at the governance boundary.

Deployment behaviour is governed by the agreed configuration and contract.

In the public score notation, A means Accuracy, R means Risk, E means Ethical Compliance and S is the composite score.

Public materials do not disclose production thresholds, weights or RulePack internals.

Ask Mergetic is a deterministic, non-generative website interface. It matches a visitor's question locally in the browser to fixed, pre-approved answers, verifies the cited entries against the bundled public claims register, and refuses when no registered answer matches.

Questions are processed locally and are not transmitted by the interface. Transmission occurs only if the visitor deliberately uses an email or booking link; future analytics must not collect question content.

Mergetic is designed to deploy as middleware at an agreed API or release boundary so organisations can add governance without replacing the underlying model logic.

The final integration pattern, data flow, latency budget and required engineering work are deployment-specific and must be validated during scoping.

The current research baseline comprises 312 governed decisions in the Stage-2 governance benchmark.

Research configuration; not a production deployment figure.

In the current research configuration, the binding governance layer accounts for under one-fifth of runtime.

A research measurement, not a service-level commitment. Absolute timings are configuration-dependent and controlled.

In the current research configuration, the binding governance layer adds cents, not euros, per governed decision.

A research measurement, not a production price or cost commitment. Absolute cost figures are controlled.

Absolute cost and timing figures, rule-level governance traces, thresholds, weights and configuration provenance exist in a controlled operational profile.

The existence of controlled material is public; its values and contents are not. Do not infer or estimate them.

The Mergetic Reliability Benchmark comprises more than 100 items across more than 25 regulatory frameworks, with documented gold-standard and predict-first methodology.

A constructed research dataset. Reliability validation is reported separately and no predictive-performance claim is made here.

Mergetic has completed an end-to-end governed pipeline proof-run on an EU model family across 99 items, producing complete evidence records.

Existence claim only. It is not a claim of model equivalence, production performance, service level or published per-item cost.

The Mergetic Reliability Benchmark and Mergetic runtime governance are distinct offerings. The Benchmark provides a bounded evaluation and evidence report; runtime governance applies an active RulePack at an agreed operational boundary. An organisation may evaluate them separately or adopt them in sequence.

Benchmark pricing is not published. A benchmark result is not a production-governance deployment or a claim of production performance.

Independent inter-rater validation of the Reliability Benchmark methodology is in progress with the Applied Innovation Unit, University of Galway, under Enterprise Ireland Innovation Voucher IV20250487.

Validation is in progress. No completed validation outcome or endorsement is claimed.

The Mergetic architecture is the subject of Irish patent application IE 2025/0516 and a corresponding EPO filing, EP 26175745.4, completed on 30 April 2026. Mergetic is an EUIPO-registered trademark, number 019273452, in Classes 9 and 42.

Patent applications are pending and do not imply that a patent has been granted.

The founding Governed Workflow Pilot is €25,000 plus VAT for eight weeks, one bounded non-production workflow, an agreed evidence set normally covering approximately 250 to 500 governed decisions, and an executive readout.

Subject to fit, final scope and signed contract. The final decision range is agreed during scoping.

€15,000 of the founding pilot fee is credited toward the customer’s Early Production licence when the parties proceed under the applicable offer terms.

Subject to the signed pilot and production terms.

Annual production licensing is scoped by governed workflows and decision exposure rather than seats, tokens or generic model calls. Capacity is sized from the measured pilot operating profile and the final price is agreed before pilot completion.

Annual production figures are not published and must not be estimated from the pilot price.

The founding pilot offer is limited to three organisations.

Availability can change; verify the offer before relying on it.

The EU AI Act applies progressively. Article 50 transparency duties generally began applying on 2 August 2026; a transition to 2 December 2026 applies to specific pre-existing systems covered by Article 50(2). Regulation (EU) 2026/1744 deferred the Chapter III high-risk requirements for Annex III systems to 2 December 2027 and Annex I systems to 2 August 2028. Article 72 was not included in those Chapter III deferrals.

General regulatory information only. Applicability depends on the system, role and facts; this is not legal advice.

PRA SS1/23 took effect on 17 May 2024 and sets model-risk-management expectations for specified UK banks, building societies and PRA-designated investment firms with internal model approval. Its five principles include governance and independent model validation, and it identifies AI and machine-learning model risk within broader model risk. A clarifying amendment took effect on 23 April 2026.

The April 2026 change was a clarifying amendment, not the start of the original policy. Mergetic can support evidence and independent challenge; it does not replace a firm’s model-validation framework or establish compliance.

Mergetic publishes bounded capability mappings for EU AI Act Article 9 risk management, Article 12 record-keeping, Article 14 human oversight, Article 50 transparency and Article 72 post-market monitoring, describing where runtime evaluation, decision-time evidence, disclosure checks and human-review routing can support operationalisation.

Capability mapping only. It is not legal advice, certification, a conformity assessment or a determination that a deployment complies with the EU AI Act.

For configured outward-facing text workflows, Mergetic can verify at the release boundary that a required AI-transparency disclosure relevant to EU AI Act Article 50 is present. A failed check can require REGEN or prevent release pending review, and the disposition and reasons record whether the check passed.

Text-output capability support only. Mergetic is not a synthetic-media marking or watermarking tool, and this is not a compliance determination.

In the public website demonstration and Ask Mergetic interface, sample material and question text are processed locally in the visitor's browser and are not transmitted by those interfaces. Contact or pilot-planner details leave the browser only when the visitor deliberately submits them.

This describes the public website interfaces, not a production deployment. Production data flows, residency, retention and subprocessors are agreed during deployment and contract scoping.

Mergetic is developed by Gratitude Beacons Ltd, CRO 798751, an Irish company. Its founder is John Walsh, ACCA.

Organisational and accountable-contact identification only.

Mergetic provides governance and evidence capabilities. It does not provide legal advice, certify compliance with a law or regulatory framework, or represent illustrative demonstration data as a production record.

Standing limitation applying to all Mergetic public materials.