A
AccuFind Virtual Assistant

AccuFind Virtual Assistant answers questions from Australian and New Zealand law and a government agency's own policies at once. AccuFind Pty Ltd, an Australian owned company founded in Adelaide, sells it to government legal teams and frontline officers, and the product was previously called AccuFind Law. Each answer shows which passages come from legislation and which from internal policy, drawing on 801,178 primary source documents: consolidated law for the Commonwealth, states, territories and New Zealand, and full text judgments from 68 courts and tribunals.

A deeper research mode shows its retrieval steps and the documents it consulted. A second model then checks each answer against its cited sources, and a separate check confirms the citations point to supplied material, giving a Pass, Review advised or Fail status while leaving the decision with the officer. Administrators choose which policy files the workspace holds.

A state police pilot drew 111 officers and staff, 1,018 questions and 85% satisfaction, and an experienced sergeant reached an evidenced answer in about a quarter of the usual time. The service runs on AWS in Sydney, model inference included, and is quoted, with pilots on an agency's own documents.

Capability grades

All 15 axes, graded from public sources on the date shown. Hover a grade to see what the letter means on that axis.

AA on AI CentralityThe artificial intelligence is the product. Remove the models and there is nothing left to sell.

AI Centrality

How much of the product is actually AI. Whether the machine learning is the mechanism the buyer is paying for or a feature layered onto conventional software, and whether the vendor is specific about which is which.

The Virtual Assistant uses models to turn agency questions into responses grounded in policy and law. A deeper mode takes additional retrieval steps and shows the resulting trail. A second model checks a generated response against the question and the cited passages. Administrators choose the policy documents and the user interprets the results. The underlying primary law and agency libraries support the assistant, but the subscription is sold around model generated research and policy answers. Without their generation and checking, the named product would lose the workflow it is built to deliver.

Source: Vendor Published
BB on Citation Accuracy and Hallucination DisclosureGrounding is real and documented, with linked primary sources and a described retrieval method, short of published accuracy figures an outsider can test.

Citation Accuracy and Hallucination Disclosure

Whether the vendor publishes measured accuracy on citations and assertions, grounds output to primary sources, and says plainly what its system does when it does not know. Legal has a documented public record of fabricated citations reaching filed briefs, so an untested claim of accuracy is not evidence.

Answers separate legal sources from the agency's own policy and open the exact passage behind each source marker. One check confirms the markers point to sources the system actually received, and a second model judges how fully and accurately the answer responds, with a missing citation producing a Fail label while the answer stays visible. AccuFind's evaluation method and an expert reviewed trial on a fixed question set are on record.

It has released no test set or error rate covering the current corpus of 801,178 documents, and no automatic check of later case treatment.

Source: Vendor Published
AA on Autonomy and Oversight ModelWhat the system runs alone, what constrains it, and how a lawyer checks it are all published: modes, thresholds, review surfaces, and the route a matter takes back to human judgment. A categorical limit on a named mode or tier, stating what its output may not be used for, meets the threshold limb without a number.

Autonomy and Oversight Model

What the system decides on its own, what a lawyer must approve, and whether the vendor documents where the review point sits. A tool that drafts under review and a tool that files without one are different products and different risks.

Staff can choose an ordinary response or use deeper research, which shows retrieval steps and the documents consulted. The answer presents source passages and an advisory status of Pass, Review advised or Fail. A response with no citations receives Fail. The checks leave the response visible and do not change it. The terms disallow decisions with legal effect made solely by automation and require human review. A staff member can open source passages to verify the result before using it. Administrators control which policy documents are made available for answers.

Source: Vendor Published
BB on Operational and Outcome EvidenceReal deployment evidence with substance, short of full attribution or measurement: a named customer without figures, or figures without the named customer.

Operational and Outcome Evidence

Named, dated evidence that the product works in production at real firms or legal departments. Case studies with figures and identified customers count. Unattributed testimonials and launch announcements do not.

AccuFind reports a state police pilot with 111 officers and staff, 1,018 questions and 85% satisfaction, in which an experienced sergeant reached an evidenced answer in about a quarter of the time. An earlier trial with South Australia Police was announced for 100 frontline officers. A practice guidance project ran four rounds of expert review on a fixed question set and then a two week practitioner trial, and nine in 10 participants used the assistant without being required to.

A state government legal office also tested it with its legal staff. The completed pilot figures do not name the agency, and none of the results measures the quality of the decisions officers made.

Source: Vendor Published
BB on Privilege and Confidentiality PostureSubstantive published commitments on confidentiality and training use, short of the full picture: commonly silence on segregation between users or matters, or on what the underlying model provider may retain.

Privilege and Confidentiality Posture

How client confidences are handled: attorney client privilege and work product treatment, segregation of one client matter from another, whether client data trains any model, and what the vendor commits to in writing rather than in marketing.

The terms bar using submitted requests, documents and responses for model training and limit their use to the service. The privacy policy isolates tenants at the application, database and infrastructure layers, and detected personal information is redacted before model processing. Agency administrators manage policy files, and staff sign in with the agency's identity provider. Research prompts and responses stay in audit logs for seven years, apart from the export and deletion schedule for account data.

Model providers and hosting are named, but privileged communications get no policy of their own and no matter walls sit inside an agency's tenant.

Source: Vendor Published
BB on UPL and Professional Responsibility PostureA real position is published on advice versus tooling, short of full treatment: commonly a disclaimer without the supervision and competence dimension, or silence on jurisdiction limits.

UPL and Professional Responsibility Posture

Whether the vendor is clear that it supplies a tool rather than legal advice, who its audience is, and how it addresses unauthorized practice of law, competence and supervision duties, and jurisdiction limits. ABA Formal Opinion 512 is the reference point. Where the advice line is not the duty a product raises, the axis is read through the nearest professional duty it does raise: judicial conduct rules and the reviewing duty for products sold only to courts, and the duty to bill for time actually spent for products that draft time entries.

The product supports government legal staff and frontline officers who must apply local law and agency policy, under the line that its products suggest and officers decide. The terms separate the service from legal advice, bar presenting AI output as personal professional advice and require human review. An officer can examine the source passage without handing a legal decision to the model. Jurisdiction coverage is stated, and a regional policing AI framework applies. No bar opinion is tied to the workflow, and no procedure sets how lawyers supervise frontline use.

Source: Vendor Published
BB on AI Governance and Bias DisclosureA published governance framework with real substance, short of testing results or a named owner.

AI Governance and Bias Disclosure

Published governance over model behavior: who owns it inside the vendor, what is tested before release, and what is disclosed about disparate output across matter types, parties, or populations.

AccuFind's AI policy applies the nine AI principles of the Australia New Zealand Policing Advisory Agency, covering accountability, privacy and human oversight. Its whitepapers describe a method for evaluating grounded answers that separates retrieval failures from generation failures in expert testing, and another on taking AI into production. A government practice guidance trial used fixed questions and four rounds of expert review before practitioners got access.

The answer status tells users whether the checking model found concerns. AccuFind has released no measure of uneven performance across populations and no results by legal category for the full product.

Source: Vendor Published
AA on AI Safety and Data StewardshipRetention, deletion, access control, subprocessors and incident practice are all published, current, and specific enough to hold the vendor to.

AI Safety and Data Stewardship

Retention, deletion, access control, and what happens to prompts and documents after they are processed. Whether the vendor states its subprocessors and its incident practice, or leaves the buyer to assume.

The privacy policy covers tenant isolation, encryption and access controls. The terms identify infrastructure and model providers and set notice at least 30 days before adding a subprocessor. Audit logs containing prompts and responses are retained for seven years. A departing organization can export available data within 90 days, after which deletion applies apart from retained audit logs. AccuFind commits to notifying an affected organization within 72 hours of confirming a breach. It also says detected personal information in research requests is automatically redacted before model processing.

Source: Vendor Published
BB on AI Liability and RecourseA real published position on liability, short of the full picture: commonly a stated indemnity without scope or caps.

AI Liability and Recourse

What the vendor stands behind contractually when its output is wrong. Indemnities, caps, carve outs, insurance, and whether any of it is published or only reachable through a negotiated agreement.

Losses from relying on AI output without independent professional verification are excluded. General liability is capped at the fees paid in the 12 months before the claim, with a separate cap of three times those fees for breaches of data security or confidentiality. An organization can terminate for a material service failure left unremedied after notice, and must indemnify AccuFind in specified circumstances. No general indemnity or accuracy warranty covers conclusions the assistant draws from legislation or policy.

Source: Vendor Published
BB on Practice Systems Integration DepthReal integrations exist and are documented, short of depth: named connections without a description of what they actually move.

Practice Systems Integration Depth

How deeply the product reaches into the systems legal work already lives in: document management such as iManage and NetDocuments, Word and Outlook, contract lifecycle management, matter management, e-billing, and court filing systems.

Staff sign in through the agency's identity provider, and an administrator adds, replaces or removes agency policy files, each screened for security before it can feed answers. The assistant retrieves those files together with the indexed legal sources, about 7 billion tokens in all. After termination the data terms give a machine readable export route. No live connection to a named legal document or case management system syncs content and permissions in either direction, so integration covers identity and administered agency content.

Source: Vendor Published
AA on Deployment Model and Data ResidencyDeployment options and data residency are published, including the regions available, what changes between tiers, and where processing happens as distinct from where data is stored.

Deployment Model and Data Residency

Where the software runs and where the data sits. Multi tenant cloud, single tenant, private deployment, on premises, and whether region of residence is a published option or an enterprise conversation.

Hosting sits in the AWS Sydney region, with model inference through regional Amazon Bedrock endpoints, and PolicyDNA and Atlas run there too. Regional service controls keep data from leaving Sydney for products hosted in Australia, covering both storage and model processing. Each agency's content is isolated under the tenant design. Redact, the company's separate redaction product, can also be hosted in the customer's own country or deployed into its own AWS account, but the Virtual Assistant runs only under the Sydney arrangement, with no second region on offer.

Source: Vendor Published
DD on Security Certifications and Trust CenterNo independent security attestation located.

Security Certifications and Trust Center

Independent attestation a buyer can pull without a sales call: SOC 2, ISO 27001, penetration test summaries, a trust center with current reports and named scope rather than a badge image.

Encryption, tenant isolation, staff access controls, security assessments and penetration tests protect the service. Agency staff sign in through their identity provider, policy files are screened before the assistant can use them, data stays in Sydney, and a breach is notified within 72 hours of confirmation. AccuFind holds no current independent certification or assessment scoped to the Virtual Assistant, with no report period and no report to request, and the controls of the AWS services underneath do not attest to AccuFind itself.

Source: Operator Verified
BB on Model Supply Chain DisclosureThe supply chain is partly disclosed: providers named without change notification, or architecture described without the providers.

Model Supply Chain Disclosure

Which models sit underneath, whose they are, where they run, and whether the vendor commits to telling customers when that changes. A legal buyer inherits every dependency it cannot see.

The service terms effective 6 December 2026 name AWS as the only subprocessor, for infrastructure and model inference through Amazon Bedrock and Amazon SageMaker, and state that the companies making the models receive no client data. The privacy policy places all AI models in the AWS Sydney region and adds that Bedrock's terms bar training on customer inputs and outputs. A second model checks answers, but neither it nor the main model is named for each mode.

AccuFind gives advance notice of any new subprocessor. No matching promise covers a change of model within an existing provider, and no current map ties models to tasks.

Source: Vendor Published
DD on Commercial TransparencyNo pricing information published at any level, including the unit of charge.

Commercial Transparency

Whether a buyer can learn what this costs without entering a sales process: published rates, the unit being charged, what sits behind an enterprise tier, and what implementation adds.

Agencies buy the Virtual Assistant by subscription. An agency can first run a pilot on selected documents alongside its current process and review the security setup before wider use. The terms leave the subscription period, payment schedule and scope to each customer's agreement. No standard amount, billing unit or minimum exists for the Virtual Assistant, and onboarding, document preparation and pilot costs carry no public schedule.

AccuFind's other products, Redact, PolicyDNA, RequirementsDNA and Atlas, are priced separately and set no rate for this subscription.

Source: Vendor Published
BB on Firm and Practice CoverageSegment and practice coverage is described with substance, short of the boundaries: what is supported is clear, what is not is left open.

Firm and Practice Coverage

Who the product is actually built for. AmLaw, midlaw, small firm and solo, in house departments, government and courts, and which practice areas are supported rather than merely claimed.

AccuFind sells to government legal, policy, freedom of information and frontline teams in Australia and overseas. Its examples include a state government legal office, a state police force and practitioners consulting statutory guidance, and it points to one agency whose operational policy grew from about 300 pages in 2006 to 7,000 in 2026. The product joins agency policies with legislation for the Commonwealth, the states, the territories and New Zealand and with judgments from 68 courts and tribunals, and administrators pick the internal material. Law firm practices and packages by organization size are not described.

Source: Vendor Published
Sources on file

6 public documents

The public pages on file for AccuFind Virtual Assistant, with the recorded signals each one supports and the date it was last read. Open any of them and check the reading against the record.

Pricing

No published figure

  • The Virtual Assistant is sold to an organization by subscription, with the period, scope and payment dates agreed with each agency. An agency can first pilot it on its own documents alongside its current process. No standard rate or billing unit is listed.

The organization negotiates its subscription. The agreement sets subscription periods, scope and payment schedules. The vendor offers an agency pilot and quotes wider deployment.

Implementation: The pilot runs on an agency's own documents with a security review, and no pilot or implementation fee is listed.

Note: AccuFind's other products, Redact, PolicyDNA, RequirementsDNA and Atlas, have their own scopes and set no price for the Virtual Assistant.

Legal Signals

What each signal means

A signal records what public sources say on the date shown. It is not a grade and it is not a recommendation. Where a signal reads Not addressed, it means the index did not locate the material in public sources on that date, which is a statement about disclosure rather than about the product.

Confidentiality and Privilege

Client Data in Training

Can material a lawyer puts into this product be used to train a model?

Never, in the contract

The published terms prohibit training on customer content. Not a policy page, the agreement.

The terms expressly bar training models on research requests, responses and documents, and limit their use to running the service. A separate clause allows deidentified, aggregated monitoring of the service without sharing research patterns with outside parties.

Source: Vendor Publishednot for model trainingAs of Nov 20, 2026Evidence

Prompt and Output Retention

How long does the product keep what a lawyer typed, and can that be set to zero?

Disclosed fixed window

A specific retention period is published and the customer cannot change it.

Audit logs holding request and response text are kept for seven years. After termination the organization has 90 days to export its data, and deletion follows for everything except logs still inside that retention schedule. Detected personal information is redacted before a request reaches the model.

Source: Vendor PublishedAudit logs (queries and responses)As of Nov 20, 2026Evidence

Ethical Walls and Matter Segregation

Does retrieval respect the firm’s ethical walls, or can the model read across them?

Own model, documented

The product maintains its own permission model, documented, requiring the firm to keep it aligned.

Tenants are isolated at the application, database and infrastructure layers, and each agency controls its own policy files and staff identity. Inside one agency's tenant there is no separate wall between matters or teams.

Source: Vendor Publishedenforced at the application, database and infrastructure layersAs of Nov 20, 2026Evidence

Third Party Request and Subpoena Notice

If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?

Disclosure addressed, notice absent

Published terms or policy address disclosure to authorities or in response to legal process, and no commitment or reservation regarding customer notice is located anywhere. The vendor has told the customer that data can leave and has said nothing about whether the customer hears of it.

The privacy policy allows disclosure where the law or valid legal process requires it. The terms promise notice of breaches and of new subprocessors, but not of an authority's request for an agency's data.

Source: Vendor Publishedif the law or a valid legal process requires itAs of Nov 20, 2026Evidence
Accuracy and Authority

Primary Law Corpus Provenance

Where does the law in this product come from, and does the vendor have the right to use it?

Sources named, basis unstated

Sources are identified without stating the license or rights basis.

The collection holds 801,178 primary source documents: consolidated legislation for 10 jurisdictions, full text judgments from 68 named courts and tribunals and the Australian Taxation Office Legal Database. The coverage catalog gives update timing for each collection but no rights basis for any of them.

Source: Vendor PublishedAs of Nov 20, 2026Evidence

Good Law Verification

Does the product tell you when the authority it just cited has been overruled?

Not addressed

No located public material addresses whether authority is checked for subsequent history.

Each answer sources its passages from current consolidated law and reported judgments, and the checks confirm that citations exist and support the answer. Nothing calculates how later courts treated a cited case or flags one that has been overturned.

Source: Vendor PublishedAs of Nov 20, 2026Evidence

Refusal and Uncertainty Behavior

What does the product do when the answer is not in the corpus?

Confidence signal only

The product exposes a confidence or grounding score without an explicit abstention path.

A second model marks each answer Pass, Review advised or Fail, and an uncited answer always gets Fail. The status is advisory, so the answer stays visible and unchanged, and the assistant does not withhold one.

Source: Vendor PublishedAs of Nov 20, 2026Evidence

Fabricated Citation Record

Does a public court record exist addressing fabricated or hallucinated legal citations in output from this product?

None located

No court order, opinion or disciplinary record addressing fabricated or hallucinated legal citations produced by this product has been located as of the date shown. This is a statement about the public record on that one subject, not a finding about the product, and this signal is not a litigation history.

The AI Hallucination Cases database kept by Damien Charlotin has no entry naming AccuFind Virtual Assistant, AccuFind Law or Legal Oracle. Every answer cites its passage, and an uncited answer is marked Fail.

Source: Operator VerifiedAs of Nov 20, 2026Evidence
Professional Responsibility

Bar Guidance Alignment

Has the vendor engaged in public with the ethics opinions its buyers are bound by?

Generic reference

Public materials refer to professional responsibility in general terms without naming guidance.

The AI policy follows the Australia New Zealand Policing Advisory Agency's nine principles and keeps final decisions with officers and lawyers. It cites no professional conduct ruling from an Australian law society or bar association, and no named ethics opinion.

Source: Vendor PublishedAs of Nov 20, 2026Evidence

Billing and Fee Posture

Does the vendor address what happens to the bill when the work takes an hour instead of six?

Outside the fee relationship

The product does not touch a fee between a lawyer and a client. It operates before an engagement exists, or it is bought by a team that bills no client for the work. Savings claims aimed at the buyer’s own cost are recorded in the summary and do not make the row a savings claim, because no client bill is in the loop.

Government legal offices, police forces and policy teams buy the service for their own internal work. They bill no outside client, and the subscription and pilot fees sit with the agency.

Source: Vendor PublishedAs of Nov 20, 2026Evidence

Outside Counsel Guideline Readiness

Can a firm get this vendor through a client’s AI clause without a bespoke negotiation?

Subprocessors listed

A current subprocessor or model provider list is published.

The terms name AWS as the subprocessor for infrastructure and model inference through Amazon Bedrock and Amazon SageMaker in the Sydney region, and promise at least 30 days' notice before a new subprocessor is added. Breaches are notified within 72 hours. There is no client facing disclosure package and no independent security certification.

Source: Vendor PublishedAs of Nov 20, 2026Evidence

Court Disclosure Support

If a judge’s standing order requires an AI disclosure, can the product produce one?

Partial record

Some elements of the record are available, short of a document level export.

Cited passages and the steps of a deeper research run can be inspected, each answer carries its review status, and audit logs keep every request and response for seven years. A departing organization can export prompts, outputs and logs. No export names the exact model and records a lawyer's verification for a court.

Source: Vendor PublishedAs of Nov 20, 2026Evidence
Contact

Correct a record, or ask how something was graded

Every grade and every signal on this index is drawn from public sources and dated. If a record is wrong, out of date, or missing an artifact the index did not locate, send the source and it will be reviewed and the record redated. Vendors are welcome to submit documentation. Nothing on this index is for sale, including a listing, a placement, or a grade.

AI Legal Index

The AI Legal Index is an independent index that tracks changes to AI vendors in legal. It holds 617 vendors across 9 categories, each graded on the same 15 capability axes and recorded against 12 legal signals, from public sources. No vendor pays for inclusion, placement, or rating.

Index Status
Last index update
November 20, 2026
The AI Legal Index is an editorial reference. It is not a regulatory body, not a law firm, and nothing published here is legal advice or a recommendation to retain or avoid a vendor. Records are verified against published sources, bar guidance and public court records. Where a record reads not addressed, the material was not located in public sources on the date shown. See the Methodology page for evaluation standards and limitations.
© 2026 AI Legal Index
3801 N Capital of Texas Hwy, Ste E240 · Austin, TX 78746