Particula

The removal test leaves a real product behind, which is why this sits below the category norm. Machine learning carries the continuous layer: the monitoring platform evaluates more than ninety smart contract metrics and ingests validator distribution, wallet concentration and market liquidity across networks, refreshing on chain inputs in near real time and updating scores automatically.

The initiating rating is analyst led, with expert analysts performing due diligence against a published framework and a formal review at least quarterly. Strip the models and quarterly analyst reports on a few hundred assets remain, which is a weaker but functioning ratings business of the kind the established agencies have run for a century. This is the human curation plus machine throughput shape already recorded for regulatory intelligence vendors in this index, and it earns the same grade for the same reason.

Oversight of the rating production path is described in unusual detail and oversight of the enforcement path is not described at all. On the production side, all incoming data is validated, cross verified and plausibility tested before it informs an assessment, analysts perform the initial due diligence, and every score undergoes a formal review at least quarterly with automated updates in between. Those are named controls at stated positions in the workflow.

The gap sits on the other side of the product. The Digital Asset Risk Passport is built so that protocols can query and enforce ratings in real time, which means a score movement generated automatically between quarterly reviews can trigger an on chain consequence with no person in that loop. Nothing states whether a materiality threshold, a delay, or a review gate stands between an automated score change and its enforcement.

Framework transparency is a real strength and it is not transparency about the models. The Particula Digital Asset Risk Framework is published, the rating process is described stage by stage from scoping through due diligence to continuous monitoring, update cadence is specified by data type, and individual ratings are released as reports carrying their rationale and supporting data.

An institution can therefore reconstruct how a judgement was reached, which is the material a validation function actually needs and which most of this index withholds. What is absent is any account of the machine learning layer itself: no model description, no training or validation approach, no back testing, no calibration evidence and no artificial intelligence management system certification. This is the distinction already recorded in this index against a model authoring platform, where the transparent thing and the modelled thing were not the same thing.

The evidence is strong on the dimension that matters most for a ratings business, which is checkable public output. Individual ratings are released as full reports with rationale, supporting data and dates, including a triple A assignment to the Janus Henderson Anemoy collateralised loan obligation fund token in November 2025, an upgrade of a Delta Wellington ultra short treasury fund token from double A plus to triple A in September 2025, and a pre issuance B plus on a credit enhanced digital loan note in March 2026.

That is a falsifiable public record rather than a claim about one. Named counterparties arrive with named executives attached: the chief executive of infiniFi on the underwriting mandate, the capital markets lead at INX, and the head of strategy for the digital economy at Moody's Ratings on a collaboration evaluating operational risk approaches.

Allfunds Blockchain, a distribution platform with real money moving across it, named the firm its risk assessment partner for tokenised funds on Solana. The limitation to record is that no calibration or accuracy study exists, and the claim of no defaults or security breaches in rated products to date is self reported over a short and benign period.

The dual position shape recurs here and this is the fourth instance recorded in this index. The firm receives non public documentation from issuers seeking a rating while serving allocators deciding whether to buy those issuers' instruments, and in the infiniFi engagement it operates an allocator's risk infrastructure directly while independently rating assets that may flow through it.

The partnership announcement addresses this once, stating that the arrangement covers technical and operational integration only and does not influence analytical independence, methodologies, assessment criteria or rating determinations. That is more than most vendors say and it is an assertion rather than a described separation.

Nothing states what information barrier exists, whether one issuer's submitted material can inform the scoring of another, or how the monitoring models are prevented from carrying non public inputs across engagements.

Personal data exposure is genuinely low, which makes the silence less consequential here than in most of this index, and the silence is nonetheless complete. The firm rates instruments rather than people and the bulk of its intake is public chain data. No privacy policy, data processing agreement, subprocessor list or retention schedule was located.

The material that would matter is issuer supplied: the methodology describes secure data collection from project teams and the incorporation of verified issuer documentation, which is commercially sensitive non public information about a rated entity, and nothing states how long it is held, where it sits, or who inside the firm can reach it.

No certification, attestation or trust centre was located in the material reviewed. No service organisation control report, no information security management system certification and no dedicated security page surfaced across the company site, its press releases, or the independent research covering this category.

The absence is recorded as unlocated rather than proven, because this index has twice found a vendor's entire security posture on a compliance page that product and company pages never link prominently, and that possibility has not been ruled out here.

What can be stated is that a firm which describes secure data collection from project teams in its own published methodology, and which incorporates verified issuer documentation into its assessments, publishes nothing about how that material is protected.

The firm publishes letter grade ratings using the vocabulary of regulated credit ratings, including triple A, double A plus, triple B plus and B plus assignments on fund and note instruments, and it holds no credit rating agency registration.

Independent research covering this category records the three established agencies alongside it as registered under the United States nationally recognised statistical rating organisation regime and on the European securities regulator's credit rating agency register, and records no equivalent standing for this firm. Membership of a token standardisation association is an industry body affiliation and carries no supervisory weight.

No sandbox participation, regulator programme admission or supervised test of the product was located. This is a factual observation rather than a legal conclusion, and it is the first thing an institution embedding these ratings in a regulated process would need to resolve.

Nothing on fairness testing, differential outcome monitoring or model bias is published, and the structural exposure here is specific rather than generic. The methodology incorporates verified issuer documentation where available, which means an issuer with a mature documentation function is better evidenced than an equally sound issuer without one, and the assigned rating reflects that difference rather than the underlying asset quality.

Rated outcomes determine which tokenised instruments reach institutional distribution and on what terms, so a systematic tilt toward well resourced issuers would shape access to capital rather than merely describe it. Nothing addresses whether ratings are tested for that effect, and nothing describes whether the scoring layer behaves differently across chains, token standards or issuance structures.

No liability position, error rate, correction path or appeal route is published, and the consequence chain is longer than usual because these ratings are built to be acted on automatically. An institution allocating against a rating, an issuer whose instrument is marked down, and a protocol that enforces a score change through the Digital Asset Risk Passport are three parties with three different exposures, and nothing states where responsibility sits among them when a rating is wrong.

The established agencies operate inside a body of law, mandated disclosure and litigation history built up over decades around precisely this question. A firm publishing the same letter grades outside that perimeter inherits none of it and offers nothing in its place.

No model provider, family, version, hosting arrangement or country of processing is disclosed. The published material refers to advanced machine learning models processing millions of on chain data points and goes no further. Because the intake combines public chain data with confidential issuer documentation, the question of whose infrastructure that documentation passes through is a live one for any issuer submitting material for a rating, and no answer is offered anywhere in the published material.

The integration surface is unusual and it is named. The Digital Asset Risk Passport, launched in March 2026, encodes ratings and their metadata as on chain objects that protocols can query and act on directly, which makes a rating machine readable at the point of allocation rather than a document a person reads.

In the infiniFi engagement the company took primary operational responsibility for the allocator's risk assessment infrastructure, including optimisation of an existing framework and its assessment questionnaires, which is materially deeper than supplying a data feed.

Held off the top grade because the conventional integration evidence is missing: no application programming interface reference was located, no integration count is published, and no portfolio, order management or fund administration system is named as supported.

No deployment options, hosting regions, tenancy model or residency commitments are published. The position is complicated by the firm's own history: it was founded in Munich and announced a relocation of its headquarters to the United States in April 2025, which is a material change for European institutional clients whose submitted documentation is processed somewhere, and no data residency statement accompanies the move.

The Digital Asset Risk Passport publishes rating output onto public blockchains, which is a deliberate and stated design choice rather than a gap, and it is the only part of the deployment picture that is described anywhere.

No pricing, tier structure or billing basis is published, which is the index norm, and a second and more consequential omission sits behind it. The firm does not state who pays for a rating. Whether an issuer commissions and funds the assessment of its own instrument, or an allocator subscribes to independent coverage, is the single structural fact that determines where the conflict sits in any ratings business, and it is the question the established agencies were eventually compelled to answer in public. The published material describes secure data collection from project teams on one side and mandates from allocators on the other, which points both ways at once and settles nothing.

Buyer breadth is genuine within a single asset class. The stated customer base spans banks, asset managers and trading facilities, and the named relationships bear that out: Allfunds Blockchain for tokenised fund distribution, the on chain allocator infiniFi, and the regulated trading venue INX, whose capital markets lead is quoted by name.

The issuers behind the rated instruments reach large traditional managers, including a Janus Henderson fund token and a Wellington treasury fund token, so the underlying counterparties are institutional even where the wrapper is new. Coverage is bounded by asset class rather than by buyer type, and more than two hundred assessments is a modest book measured against agencies rating thousands of instruments.