ArsLexStraight answers, in plain English.
Litigators are right to be careful with new tools. These answers explain what the Discovery Suite does, what the AI is and is not allowed to decide, what it costs, and how we protect your confidential work. No hype.
Getting a matter in
What can I bring in?
Mailboxes (PST, OST, mbox), loose files, zips and other archives, a drive's worth of folders, and finished productions from the other side with their DAT and OPT load files. Nothing is refused for size. Every upload is kept exactly as it arrived, so a matter can always be rebuilt from what the client actually sent. If a client or a vendor needs to get files to you, you send them a link and the files land in the matter without anyone needing an account.
What happens to the file inside the file?
It becomes its own document. A spreadsheet embedded in a Word file, a PDF attached inside another PDF, an email attached to an email: each is pulled out, kept in its family, and reviewed and produced as a document in its own right. Audio and video are transcribed. CAD drawings and Microsoft Project plans get pages you can read. A scan in another language is read again in that language. Photographs are described in words you can search. A password-protected file is held rather than dropped, and a per-matter password bank tries every password you have on record.
What if something cannot be read?
You are told. Every file that could not be processed is listed with the reason, the stage it failed at, and its hash, and that list is exportable for the meet-and-confer. A check runs at the end of every import and again each night, and reports new defects rather than quietly carrying them. Because the original upload is kept as received, you can re-import it after a fix without asking anyone for the files again.
Reviewing
What does "review by issue" mean?
You describe an issue in a sentence or two, once. ArsLex finds the documents about it and keeps finding them as new material arrives, so an issue is a standing question rather than a saved search you have to rerun. Open the issue and you read those documents in the order they happened, one at a time, with families and email threads kept whole. The server hands out the next document when you ask for it, so two colleagues working the same issue never land on the same document and never need to refresh to stay out of each other's way.
What does the AI actually do, and who decides?
The model reads each document and offers a suggestion on relevance, privilege and confidentiality, with a short summary. Its suggestion lands in its own column and stays there. Production keys off the attorney's coding only, so nothing is ever produced or withheld because a model said so. When a document has been redacted, the model reads the burned copy, never the unredacted text. And before a bulk coding call commits, you see every document where the AI reads the record differently from the call you are about to make. It is advice you can see, not a decision you did not make.
Can it code documents in bulk? What stops it from coding 5,000 of them wrong?
It can, and the answer is that you see the damage before it happens rather than after. Select a search result and ask to code it, and nothing is written yet. You get a preview first: how many documents actually change versus already carry that call, which family members are pulled along, which families would end up split with a parent produced and a child withheld, and a list of every document where the AI's read of the record disagrees with the call you are about to apply. That last list is advisory. It never blocks you, and the lawyer always wins the argument. If the gesture still turns out to be wrong, one action reverses it, and anything a colleague touched in the meantime is left alone rather than stamped over. Two things are deliberately kept out of bulk: removing a privilege mark is a one-document decision, because over-withholding shows up on a log and under-withholding is a waiver.
Do I have to read every email in a thread?
No. ArsLex works out which messages actually carry the conversation, so a reply that quotes everything above it makes the earlier copies redundant. You review the messages that carry content, and the rest of the thread follows your decision with a record of where that decision came from, rather than being quietly filled in. Attachments stay with their parent email throughout, including inside one continuous Bates range.
Who is Al?
Al is the assistant inside the review room. Ask it how a feature works and it answers from the product manual, with no case data involved. Ask it where something is in the production, what a term hits, or what a photograph shows, and it navigates the record for you without changing anything: it is read-only by construction. It can also run a census of the production, such as how many documents have unreadable text or are blank, and tell you before opposing counsel does. Every number Al reports is computed by the code, never by the model, and the conversation is not saved anywhere.
Producing, serving and trial
What comes out the other end?
A production the other side can load: searchable PDFs with DAT and OPT load files in the formats Relativity and Concordance read, Bates numbers that keep families in one range, a gap check before you certify, redactions burned into the page rather than drawn over it, and a privilege log written from the record rather than typed from memory. Every produced file is hashed and listed in the production index, so what you served can be proven months later.
How do I get it to opposing counsel?
Through a delivery link, optionally as an encrypted archive, rather than a courier or an FTP account. Mark the set served and it freezes: nothing in it can be edited, re-stamped or deleted, so the set still matches its certificate a year on. Your own copy stays live and searchable for the life of the matter.
What about the production I receive from the other side?
It gets the same treatment yours does. Load it with its load files, review it by issue, and log the other party's privilege claims against what they actually withheld. If they send a clawback notice, enter the Bates range and the matching documents leave review and export at once, with a record of when and why.
Can I take it to trial from here?
Yes. Mark documents as exhibits, label them to the court's scheme, print coversheets, and package them for filing. A filed exhibit set is immutable, so what was filed is what stays on record. No separate trial product and no re-upload.
Pricing and running it
How is the Discovery Suite priced?
One rate per gigabyte, per matter, per month, with AI review inside it. No seats, no per-document charges, and producing costs nothing extra. Hosting, redaction, privilege logs, productions and backups are all in the same number. A finished production with load files needs no further processing; a raw collection such as a mailbox or a drive is processed into a reviewable production and the quote says so up front. A closed matter can move to a lower archive rate and come back when you need it. Tell us the matter's size and what form the data is in, and you will have a number the same day.
Do we need a vendor to run it?
No, and you can keep one. The platform is built so a paralegal can run intake, review and production end to end without a project manager in the loop, and the hours your own people spend are your firm's to bill. Many firms prefer to have a service provider they already trust run the matter on ArsLex instead, and the platform works the same way for them.
How big a matter can it handle?
There is no size ceiling at the door, and processing runs on a fleet of workers that grows with the job rather than a single server. Productions in the hundreds of thousands of documents are in review today. If you have something unusually large, say so in the quote request and we will tell you what to expect.
The AI, plainly
What is an LLM, and why does it "hallucinate"?
A large language model is trained on enormous amounts of text to predict the most likely next words. That makes it fluent at reading, summarizing and classifying, but it is a language engine, not a fact database: it produces text that is plausible rather than verified, which is how a chatbot can invent a case citation that does not exist. Courts have sanctioned attorneys who filed such citations. ArsLex uses the model only over documents you loaded, never asks it to recall anything from memory, and never lets its output be the final word on anything that reaches a production.
Which parts are AI and which parts are rules?
Anything that must be exact is rule-based and repeatable: Bates numbering, family ranges, hashes, load files, the privilege log's fields, retention and deletion. The model is used where judgment helps and a human will check it: suggested coding, summaries, descriptions of photographs, Al's answers about where things are. The split is deliberate. The same input to a rule gives the same output every time; a model can give a different answer to the same question, so it is never placed where that would matter.
Can the AI be wrong?
Yes, and the product is built on that assumption. Suggestions are shown beside the attorney's call rather than in its place, disagreements are surfaced before a bulk action, and quality control sampling draws a statistically stated sample so the check can be reproduced. A model being wrong costs you a second look at one document; it never costs you a production.
Your role and professional responsibility
Does ArsLex give legal advice or "practice law"? Who is responsible for the work?
No. ArsLex is a tool the attorney directs, the same way a review platform or a word processor is. It does not give legal advice and does not exercise legal judgment. The attorney remains fully responsible for the coding, the privilege calls, the production and anything filed.
What about my ethical duties: competence, confidentiality, disclosing AI to the court?
ArsLex is built to support them. On competence (ABA Model Rule 1.1, generally understood to include understanding the technology you use), every AI suggestion is visible, attributable and checkable against the document. On confidentiality (Rule 1.6), client data is never used to train AI, stays on infrastructure we control, is kept until you delete it, and can be placed under a preservation hold whenever a matter requires it. On disclosure: some courts have standing orders about disclosing AI assistance, and those rules vary by jurisdiction and judge, so check the ones that govern your matter. None of this replaces your professional judgment; it is designed to help you stay on the right side of it.
Your data and your confidentiality
Where does my data go — and is it used to train AI?
No. Your documents are never used to train any AI model — ours or anyone else's, and no model provider retains them. Text generation runs on Amazon Bedrock under AWS's zero-data-retention and zero-operator-access terms. Embeddings never go to a third party at all: wherever ArsLex uses them, including search by concept and grouping a production by subject in the Discovery Suite, they are computed on infrastructure we control and never leave it. Everything is encrypted in transit and at rest, isolated per tenant, and deletable at any time. See how we handle your data for the full detail.
How long do you keep my documents? Do you ever delete them on a schedule?
We keep your work until you delete it. Your documents and work product are never deleted on a schedule, and no setting turns a deletion timer on. Only temporary copies we make to deliver or process them expire.
When you do delete something, it is gone: the record, the search index, and the file are removed from live systems immediately, and the data leaves our encrypted backups within 30 days. There is no trash can and no restore — so keep your own copy of anything you cannot afford to lose. And while your firm's agreement with ArsLex is in effect, nothing in a matter under a preservation hold can be deleted until your firm releases the hold. If your firm's agreement with ArsLex ends, you have 30 days to export your data. After that we delete it, held matters and served productions included, unless the law requires us to keep it.
What happens to my chat conversations with the assistant?
They are never saved. Conversations are held only in server memory so your signed-in devices (say, Word and your phone) can share one live thread, and they are wiped when your session ends, when you log out, or when you clear them: always within hours, never written to a database or disk, and never included in backups. Your documents and drafts are a different thing — those we keep until you delete them; the chat itself simply evaporates.
Who holds the encryption key, and can you prove a deletion actually happened?
Your documents are encrypted at rest as a matter of course. Beyond that, a firm can ask to have its documents in our durable store sealed under an encryption key that belongs to that firm alone, held in AWS Key Management Service and used for nothing and nobody else. It matters because of what it makes possible: destroying that key makes the firm's stored documents unreadable everywhere at once, which is a stronger and faster statement than walking a delete through every copy.
That is a serious thing to do, so it cannot be done quietly or quickly. Every administrator on the account is warned in the app and by email. Someone has to confirm it explicitly, and a confirmation typed against a stale warning does not count. A waiting period runs before anything is destroyed, the destruction itself takes deliberate human action from us rather than a scheduled job, and if we cannot reach an administrator at all, the request simply cannot go forward. Ordinary deletion is unchanged and is not on any of these clocks: delete a document and it leaves live systems immediately, and our encrypted backups within 30 days.
Can other firms see my data? What about deployment inside our own walls?
Each firm's data is walled off from every other tenant — isolation we enforce and test continuously. Access requires authenticated login with multi-factor authentication, and the database is never exposed to the public internet. For firms that must run the application inside their own environment, ArsLex is also available for on-premises deployment — your documents, database, and search index stay on infrastructure you control, while AI generation still calls our zero-retention model provider.
Drafting with ArsLex Associate
Do you also do drafting, or only discovery?
Both. ArsLex Associate is a Microsoft Word add-in that drafts litigation briefs from your own case law, depositions and pleadings, with Bluebook citations formatted automatically. It runs on the same zero-retention foundation, and documents from a Discovery matter can be cited in a brief by Bates number.
Can it cite a case that doesn't exist?
It cites what you give it. Associate drafts only from the documents you upload, anchors each citation to a real document at a real page, and formats the citation by rule rather than letting the model compose it. Every cite-anchored passage traces back to its source so you can open the original and check it. As with any tool, you remain responsible for reviewing the finished work before it is filed.
Still have a question? Email admin@arslex.ai. We respond within one business day.