Irys, formerly Iqidis, is a legal AI platform from Iqidis, Inc. of New York, which rebranded in March 2026; the main product, Irys One, was formerly called Iqidis. It organizes legal work around matters. It researches case law against retrieved sources, with results linked to the opinions relied on, and drafts motions, memos, contracts and letters in a firm's voice. It also redlines against playbooks, analyzes a matter's documents through a knowledge graph, and runs Cite Check, which checks citations and flags those it cannot match.
Irys Workbench agents gather material from connected email and files and return finished documents, and an add in brings Irys into Microsoft Word. Most processing runs on Irys's own retrieval and orchestration layers, with named third party models called for discrete inference steps under terms that bar training and retention. It serves solo practitioners through large firms, in house teams and insurance defense firms. Plans run from a free tier and $19.99 a month to $299 per seat a month for Irys One, with Enterprise priced to scope.
Side by side on the same 15 capability axes and 12 legal signals: where the two part company, and what neither one publishes.
Other products that do the same kind of work as Irys, and what each one does.
All 15 axes, graded from public sources on the date shown. Hover a grade to see what the letter means on that axis.
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.
Irys One researches case law, drafts motions, memos and contracts in a firm's voice, redlines against playbooks, analyzes documents across a matter through a knowledge graph built from the matter's files, and checks citations; Workbench agents gather material from connected email and files and return finished documents. Matter management, the document library and electronic signature support that work; without the models there is little left to sell.
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.
Research runs against retrieved legal sources, and results link to the opinions and documents relied on. Cite Check checks identified citations against legal source databases, flags those it cannot match, and surfaces available treatment signals such as negative history. The Thorough mode checks case law citations against source as it writes, while the Quick mode is disclosed as not backed by citations. The Responsible AI page says plainly that no mode makes output correct and that Cite Check does not verify every citation.
The company has published a run of its research system against a public legal agent benchmark of 1,251 tasks, with an 83.74% criteria pass rate and a 17.75% strict pass rate on all criteria. It also published its departures from the benchmark's method, and released the code and traces. That measures agentic legal work, not the accuracy of Irys One's citations.
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.
The Quick mode is stated not to retrieve or verify authority and never to be the basis for authority cited in a filing; the Standard research mode is to be paired with Cite Check before its citations are relied on; the Thorough mode checks citations as it works. Workbench agents gather material and build documents, but emails are left in drafts and never sent without the user. Drafts arrive as tracked changes, Cite Check returns flags for an attorney to resolve, and an audit trail records what was run, when and in which mode.
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.
The site quotes Omer Er, Partner at Michelman & Robinson, lists users at Falls Counsel, Cox O'Brien and the Law Offices of Li Yang, and says more than 300 legal teams use the platform. Its outcome figures, 41% fewer hours and turnaround cut from 3.9 to 2.3 days, are attributed to two unnamed legal teams in their first weeks on the product.
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 Data Processing Addendum and Terms of Service commit that customer content is not used to train Irys's or third party models. It is kept separate for each tenant and user profile, never commingled with other customers, and deleted within 30 days of termination or earlier on request. The terms bind Irys to confidentiality, with prompt notice before any compelled disclosure. Third party models are named, used for stateless inference with retention on the provider side disabled, and bound by terms that bar training.
Administrators cannot see a user's personal workspace unless it is shared. The security page presents the platform as designed to preserve confidentiality and support privilege, while the terms leave compliance with privilege obligations to the user. Separation by matter is described on product pages, not in the agreement.
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 Terms of Service and a separate Disclaimer state that Irys gives no legal advice, creates no attorney client relationship and is an assistive tool for legal professionals. A Professional Responsibility and Ethics section requires independent judgment, review of material output including authorities, quotations and subsequent treatment, review before anything goes to a court, client or regulator, and firm policies, training and supervision for AI tools.
A Responsible AI page explains in plain language why attorney review is not optional, and an AI Rules by State guide sets out disclosure requirements, ethics guidance, sanctions and CLE rules state by state, with sources. No consumer facing surface is published.
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.
The Responsible AI page sets out what Irys does and does not claim, how its safeguards are layered and what strong firm practice looks like; policies name Legal & Security at Iqidis, Inc. as their owner, and the security page says the architecture was audited by Guidepost Solutions. No accountable individual, internal governance framework, testing before release, or finding on whether output differs across matter types, jurisdictions or practice areas is published.
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.
Canceled accounts stay exportable for up to 90 days and are then deleted within 30; the Data Processing Addendum commits to deletion within 30 days of termination or earlier on written request; users can delete data at any time. MFA is enforced for all accounts, with least privilege roles, SSO and SCIM, idle session expiry and an exportable audit trail. A subprocessor policy names every service provider and subprocessor with purpose, location and retention, and commits to 30 days' notice of changes with a right to object.
Breaches are notified within 72 hours with set contents and equivalent obligations on subprocessors, and government requests are notified where lawful and challenged if overbroad.
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.
The Terms of Service cap each party's liability at twelve months' fees or $100, whichever is greater, but lift the cap for breaches of confidentiality and for gross negligence or willful misconduct. Irys defends customers against US patent, copyright and trademark claims over the service, including output used as delivered, and the Data Processing Addendum adds an indemnity for data subject or regulator claims caused by a security incident in Irys's systems. Irys does not warrant that output is accurate or that verification features catch every issue.
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.
Irys runs inside Microsoft Word through an add-in on AppSource, and connects on the user's authorization to Google Drive, Gmail, OneDrive, Outlook and Dropbox for import and export, as listed in its subprocessor policy. Product pages also show iManage, NetDocuments, SharePoint and Slack, but its FAQ says document management integrations are in development, and what syncs and what a firm must configure is not documented.
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.
The subprocessor policy says data is processed primarily in the US and EU with regional routing on request, hosted and stored on AWS in the US, and gives each AI provider's processing location, US and EU for the main models. Product pages describe hard tenant isolation with containers for each tenant and each user, and name cloud, VPC and on premises options. What each deployment option involves, and which plans include which, is not described beyond Enterprise being priced to scope.
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.
The security page lists SOC 2 Type II, ISO 27001 (the subprocessor policy specifies ISO/IEC 27001:2022), HIPAA, GDPR and CPRA compliance, and an architecture audit by Guidepost Solutions; current reports and certificates are available to customers and prospects under NDA on request. A trust center is linked, and the security page and subprocessor policy publish no auditor, scope or report date.
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 subprocessor policy names OpenAI for inference and embeddings, Anthropic's Claude and Google's Gemini on Vertex AI for inference, Mistral AI for OCR, Perplexity AI and Linkup for research lookups and CourtListener for case law and citation lookup, each with processing location and retention, and commits to 30 days' notice of any new subprocessor or material change, with a right to object. It also describes the architecture: most requests are handled by Irys's own knowledge graph, retrieval and orchestration layers, with external models called only for discrete, stateless inference steps. Specific model versions are not listed.
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.
The pricing page publishes a free plan, Irys Lite and Irys Student at $19.99 a month, Irys One at $299 per seat a month or $3,049 a year, and Irys Teams at $329 per seat a month for 2 to 25 seats, with usage limits by plan and no metered overage on Irys One. Enterprise, for 25 or more seats, is priced to scope, and the enterprise deployment program is not priced.
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.
Pages address solo practitioners, small, midsize and large firms, in house teams and insurance defense firms, with use case pages for litigation, corporate and contracts, insurance defense, employment, mass tort and ADA compliance, and law students have their own plan. Citation checking is described against federal and state opinions. The jurisdictions, courts and practice areas it does not support are not stated.
The public pages on file for Irys, 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.
Good Law Verification, Refusal and Uncertainty Behavior, Court Disclosure Support
Read Sep 22, 2026
Primary Law Corpus Provenance, Outside Counsel Guideline Readiness
Read Sep 22, 2026
Client Data in Training, Third Party Request and Subpoena Notice
Read Sep 22, 2026
Prompt and Output Retention, Ethical Walls and Matter Segregation
Read Sep 22, 2026
Fabricated Citation Record
Read Sep 22, 2026
Bar Guidance Alignment
Read Sep 22, 2026
Billing and Fee Posture
Read Sep 22, 2026
Published per seat subscription prices, Enterprise quoted. Free: 3 AI messages a day, 10 saved chats, 25 MB storage. Irys Lite: $19.99 a month, 12 messages a day, 175 MB. Irys Student: $19.99 a month for law students, 24 messages a day, 250 MB. Irys One: $299 per seat a month, or $3,049 a year billed annually, unlimited messages and storage, no per token or overage charges. Irys Teams: $329 per seat a month for 2 to 25 seats, adding shared workspaces, matter level collaboration, firm wide playbooks, role based permissions and admin controls.
Irys Enterprise: 25 or more seats, priced to scope, adding SSO, SAML and SCIM, custom integrations, white labeled deployment, a dedicated success manager and an MSA, DPA and procurement package. All plans include the full platform, including Cite Check and the Word add in; the free, Lite, Student and One plans are single user.
Implementation: None for the self serve plans, which start with a free trial and no setup charge. Enterprise deployments come with an onboarding program covering governance design, a rollout plan, billing strategy sessions and training; it is described but not priced.
Confidentiality and data terms: A Business Associate Agreement is available for customers subject to HIPAA, per the Privacy & Data Policy and the subprocessor policy; it is not tied to a named plan or price. The security page lists HIPAA compliance alongside SOC 2 Type II and ISO 27001.
Note: Annual billing saves up to 15%, and the page says the rate signed is the rate at renewal. The 7 day trial needs no credit card. Every plan includes 90 day data retention with automatic deletion. The Terms of Service allow fee changes on at least 30 days' notice, taking effect at the next subscription term.
Irys One Intake now accepts Slack as a source of work. Users pick public or private channels to watch, new messages land as items on the intake register with attached files brought in as documents, and thread replies join the existing item. The connection is read only, does not post or delete in Slack, and pulls no history from before setup.
Irys made a two way Clio connector generally available. Users sign in with their own Clio account, browse Clio as a matter tree, bring documents into an Irys matter or the Library for research, drafting, comparison and cite check, and ask about matter status, responsible attorney, calendar, open tasks and recent time entries. Finished drafts, notes, tasks with due dates and time entries can be written back to the matching Clio matter.
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.
Can material a lawyer puts into this product be used to train a model?
The published terms prohibit training on customer content. Not a policy page, the agreement.
The Data Processing Addendum states that customer content is not used to train Irys's own or third party models, and that third party models are used only for stateless inference with vendor caching and retention disabled under commitments not to train. The Terms of Service add that training could occur only if expressly agreed in writing. Usage data is limited to aggregated or anonymized telemetry that excludes prompts, outputs and documents.
How long does the product keep what a lawyer typed, and can that be set to zero?
The customer controls the retention window, by product configuration or by contractual instruction, but zero retention is not stated as available.
Users can delete their data at any time; Irys's pricing page says every plan includes 90 day retention with automatic deletion; canceled accounts stay exportable for up to 90 days and are then deleted within 30; and a data retention policy describing settings and default periods is available on request. Model provider inference is configured for zero retention, and telemetry excludes prompts and outputs. A setting that keeps nothing inside Irys is not described.
Does retrieval respect the firm’s ethical walls, or can the model read across them?
The product maintains its own permission model, documented, requiring the firm to keep it aligned.
Access is set in Irys's own role model, maintained by the firm: organization administrators, matter owners and contributors, with matter level team collaboration and shared workspaces per team or practice group. The terms say administrators cannot see a user's personal workspace unless it is shared, and the agreement commits to separation per tenant and per user profile. Permissions are not inherited from a document management system; those integrations are described as in development.
If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?
Terms commit to notice where lawfully permitted. No transparency report located.
The Data Processing Addendum commits Irys, where legally permitted, to notify the customer promptly of a government request for customer personal data, to challenge unlawful or overbroad demands and to disclose only the minimum required; the Terms of Service commit to prompt notice and assistance before any compelled disclosure of confidential information. No transparency report is published.
Where does the law in this product come from, and does the vendor have the right to use it?
Sources are identified without stating the license or rights basis.
The subprocessor policy names CourtListener, from the Free Law Project, for case law and citation lookup, and Perplexity AI and Linkup for legal and web research lookups; the pricing page describes citation checking against federal and state opinions. The rights basis for each source, its coverage and how often it is updated are not stated.
Does the product tell you when the authority it just cited has been overruled?
The vendor computes and surfaces subsequent history itself, with the method described.
Cite Check checks identified citations against legal source databases, flags those it cannot match, and surfaces available treatment signals such as negative history, for attorney review. Irys states that it is not connected to, or a substitute for, any third party citator a firm licenses, and the terms say a passing check does not establish that all subsequent treatment has been identified.
What does the product do when the answer is not in the corpus?
The product exposes a confidence or grounding score without an explicit abstention path.
Cite Check flags citations it cannot match against its sources, and the Quick mode is disclosed as not retrieving or verifying authority. No behavior in which the system declines to answer, or says it found no support, is described for research or drafting.
Does a public court record exist addressing fabricated or hallucinated legal citations in output from this product?
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 maintained by Damien Charlotin lists one matter naming Irys or its former name, Iqidis: in a declaration filed on 11 August 2026, counsel for a defendant states that she used the product, believed it was connected to her firm's legal research subscription, and did not check the citations before filing. No court order or finding addressing those citations is on the public record. This is a statement about the public record rather than a finding about the product.
Has the vendor engaged in public with the ethics opinions its buyers are bound by?
Public materials engage with at least one named ethics opinion.
Irys publishes an AI Rules by State guide that names the governing ethics guidance in each state covered, such as NYC Bar Formal Opinion 2024-5, Texas Ethics Opinion 705 and WSBA Advisory Opinion 2025-05, alongside disclosure rules, sanctions cases and CLE requirements, with sources. It maps state rules rather than the product's features to each opinion.
Does the vendor address what happens to the bill when the work takes an hour instead of six?
The product emits a per matter record of AI assisted work and the vendor publishes guidance on fee and disclosure treatment.
Irys publishes billing guidance covering hourly, fixed fee, value based and blended arrangements, including how to describe AI assisted work on invoices and how to answer client questions about AI use, with invoice, client disclosure and engagement letter templates provided in enterprise deployments. The platform keeps an exportable audit trail of what was run, when and in which mode. The guidance frames hourly work as freed capacity rather than addressing time billed.
Can a firm get this vendor through a client’s AI clause without a bespoke negotiation?
A current subprocessor or model provider list is published.
A current subprocessor policy lists every service provider and model provider with purpose, location and retention, and commits to 30 days' notice of changes. Audit report summaries and certificates are offered under NDA on request. No client facing disclosure material is published on the public site.
If a judge’s standing order requires an AI disclosure, can the product produce one?
Some elements of the record are available, short of a document level export.
An exportable audit trail records what was run, when and in which mode; research citations link to their sources, Cite Check flags are kept for review, and drafts arrive as tracked changes. No export for each document recording the model used, the sources retrieved and the human verification is described.
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.