We read research and development documents, map them against the rulebook that governs them — at a dated version of that rulebook — and return findings that cite the passage each one came from. Your expert still decides. The evidence trail is written for them.
Three different functions, one shape of work: read unstructured documents, apply an external rulebook, record the reasoning, decide. Then prove it years later to someone who was not in the room.
Must confirm none of them creates a biosecurity risk — before money is committed.
Must screen technical content against control lists, and prove which list version applied.
Must reconstruct its reasoning for an inspector who arrives without notice.
The cost is not the reading. It is the evidence.
Guidance issued to research funders by the Nuclear Threat Initiative instructs them to limit biosecurity review to shortlisted proposals in order to conserve resources. The organisations that need this most have already written down that they cannot afford to do it properly.
Every question in the framework is answered explicitly, and every answer quotes the sentence that drove it. A reviewer checks the quotes rather than re-reading the document.
Construction of chimeric viruses is explicitly proposed.
Sarbecovirus-derived constructs; containment level stated by the applicant corroborates.
Chimeric construction combined with passage through a humanised host is capable of enhancing a viral characteristic.
§ 5.2 describes completed work and is not assessed. Only proposed work drives the outcome.
Routed to a named assessor with the findings, the citations, and a drafted memo. Hridhaan does not approve or reject anything.
Illustrative example using publicly described policy criteria. No customer document is reproduced.
The engine does not change between use cases — only the rulebook it is pointed at, and the dated version of that rulebook in force at the moment of assessment.
Proposals, protocols, SOPs, technical files and submissions — in full, not titles and abstracts.
Against an external framework, at a specific dated version of that framework.
Structured findings, each one quoting the exact source passage it came from.
A complete audit trail: what was assessed, against which rule version, by whom, when.
To a named human reviewer, who makes the decision and signs it.
Anyone can produce a confident classification. The question a regulated buyer asks is how you measured it, and what happens when you are wrong. Each rule pack is built against an expert-labelled evaluation set drawn from the customer's own historic decisions, and no version ships until it clears the gate.
Same engine. Different rulebook. Each one produces the artifact its buyer already has to write by hand.
Screens research proposals for content that could create biosecurity or dual-use risk, before funding is committed.
Screens technical content and intangible transfer against control lists, and records the dated version applied.
Triages incoming adverse event cases and drafts the narrative, inside a validated environment.
Assembles the evidence base for a quality investigation and drafts the reasoning, on top of the existing quality system.
Checks a submission for internal contradiction across modules before it goes to the agency.
Rule packs and logic arrive as signed, versioned bundles over an outbound connection — the same pattern any licensed enterprise software uses. Where no outbound connection is permitted at all, the same bundle is delivered as a signed offline package.
Confidential proposals, patient data and trade secrets stay where they already are.
A rule pack version changes when you accept it — and every assessment records which one it ran against.
Signed offline packages where an outbound connection is not permitted.
Because it reads well. That is not the part you are being audited on.
| General assistant | Hridhaan | |
|---|---|---|
| Same input, same output | The model is upgraded underneath you, without your consent and on the vendor's schedule. | Model, prompt and rule pack are pinned and recorded. The assessment is re-runnable years later. |
| Which rulebook applied | Whatever version it happens to recall, unstated. | The dated framework version is recorded on every assessment, because that is what you will be asked. |
| Evidence | A fluent summary. Citations if you remember to ask, and not always to the right passage. | Every finding quotes the passage that triggered it, located to section and paragraph. |
| Validation | No vendor will sign that it is validated for your intended use. | Validation documentation is generated per release, against an evaluation set built from your own decisions. |
| Your systems of record | Reads. Does not write into a safety or quality system. | Writes back under formal change control, with e-signature and audit trail. |
| Accountability | Sits with whoever pasted the document in. | Contractual, and named. |
Intelligence gets cheaper with every model release. Accountability does not. Model releases commoditise intelligence, not evidence.
Four weeks. Your documents, your frameworks, inside your environment. It produces measured baseline numbers and a working prototype. Success criteria and the route to production are agreed in writing before it starts.
Book a callOr write to hello@hridhaan.ai.