Frequently asked questions.

Squire is a sovereign AI-powered legal assistant built for lawyers. It provides jurisdiction-specific legal research, drafting, and analysis using curated legal data. Unlike generic AI, Squire is grounded in your country's laws and your firm's data, and keeps your firm's work encrypted under a key unique to your firm and stored in your jurisdiction.

Squire is meant for legal professionals (attorneys, paralegals, in-house counsel) who need reliable AI legal research. It excels in complex tasks like analyzing case law, drafting documents, and finding key contract clauses. Essentially, any law firm or legal department that requires secure, accurate AI assistance benefits from Squire's specialized features.

ChatGPT, Copilot and Gemini are general-purpose assistants. Squire is built for legal work, and private. It never sends your data to public consumer AI models and never trains AI models on your content; it calls only vetted enterprise AI providers under contractual no-training terms, works from pseudonymised text, and keeps your firm's work encrypted under a key unique to your firm and stored in your jurisdiction. Its answers are grounded in curated legal sources for your jurisdiction, and its Hallucination Firewall checks citations against those sources and flags what it cannot support. In short, Squire gives you cited, source-grounded legal research and data governance that general-purpose assistants do not - and, as with any research tool, you should check the sources before relying on an answer. See how Squire compares with other legal AI options.

"Sovereign" means your firm's data, and the law Squire works from, belong to your jurisdiction. Squire's answers are grounded in your country's law, read from curated local sources and cited. Your firm's data is stored in your jurisdiction, its documents, chats, drafts and notes - and the parties and clients on its matters, its tables and what its workflows produce - are encrypted under a key unique to your firm, and before the AI analyses a document, the people in it are swapped for pseudonyms. Today that analysis may run outside your jurisdiction; we are moving AI processing onto servers we operate in your jurisdiction, starting with the step that detects personal information. On-premise deployment is available for Enterprise customers by arrangement. For a law firm this isn't just an infrastructure preference - it's an extension of privilege and the duty of confidentiality.

Squire's intelligence is built on multiple layers of legal sources, including your jurisdiction's constitution, statutes, case law, and firm-specific documents. This curated legal data (legal codes, precedents, commentary, and internal work product) ensures every answer is evidence-based. As a result, Squire cites real cases and laws rather than guessing.

The Hallucination Firewall is a core Squire feature that grounds every answer in your verified legal data and links each claim to its source, so you can check any statement against the original text. It also verifies citations against the source document and scores how well each response is supported. This is designed to keep outputs rooted in real law rather than invented - though you should always verify important findings yourself.

Squire includes document analysis (summarization, clause extraction), context-aware drafting, and deep research. It can automatically summarize long files, find relevant cases, and draft or edit documents with expert guidance. All AI tasks happen in a split-pane interface so you remain in control of edits. These features take much of the manual work out of research and drafting, while you review and decide.

Every AI-generated claim or summary is linked to its source. Squire's document intelligence engine verifies citations against the original text. You can click through to view the referenced statute or case. This transparency means you can trust the answers and double-check the details.

Yes. When you ask the assistant to produce a full document ("draft an NDA", "write a demand letter", "prepare an employment agreement"), it will offer to open a dedicated drafting session. Click "Start drafting" and Squire creates the draft, carries any research you've already done in the chat across to the new session, and opens the split-pane editor. For short snippets or quick replies the assistant will answer inline instead.

In the drafting editor the assistant can edit content, replace sections, insert new clauses, propose redlines, apply templates, save named version snapshots, restore previous versions, compare versions, change the font (Default / Times New Roman / Helvetica / Cambria), set the list-numbering style (1. / a. / i., decimal, legal 1. / (a) / (i), or outline 1.1.1) and heading capitalisation (Preserve as-typed, Title Case, Sentence case, UPPERCASE), assign a DOCX style template, rename the draft, link or unlink it from a matter, change the lifecycle status (draft / in review / finalized), and offer Word or Markdown exports. You stay in control: an edit that would delete a section or most of the text is not made until you confirm it on a card in the chat, every edit is saved as a version you can restore, and you can switch on "Review AI edits" to have the assistant propose every change for you to accept instead of making it.

Ask the assistant to export your draft and it will post a Download card in chat. Click it to download a .docx file - either clean or with your redlines converted to Word track changes. Exported documents follow your house style (font, size, spacing, heading capitalisation, and list numbering), and if your organisation has uploaded a style template they also pick up its headers, footers, margins, and branding. Markdown export is also available. For documents imported in Word review mode, choose "Word with Squire's track changes" - it returns the original document with your decisions and suggestions applied as native track changes.

Yes. Squire has a house style you control - font, body size, paragraph and line spacing, heading capitalisation (e.g. Title Case or UPPERCASE), and list numbering. Set the firm-wide default under Organization Settings -> House Style (owners and admins), and your own personal style under Account Settings -> Drafting style; your personal choices override the firm default for the drafts you create, and any single draft can still override both from the editor's Numbering and Heading case menus. The style is applied automatically in the editor, the preview, and the Word export, so documents come out looking the way your firm writes them without anyone reformatting by hand. You can also just ask the assistant - for example "make Cambria and 1.1.1 numbering my default". Numbering and capitalisation are computed at export, so they never alter the words in your document, and acronyms like POPIA or VAT are left untouched.

Yes - this is Word review mode. Upload the negotiated .docx to your document library and click "Review in Squire" (or use Import Word in the drafting editor). Squire displays the document with every existing tracked change - who made it and when - and you can accept or reject each one, ask the AI to propose changes, or suggest your own edits. Nothing is final until you export: download "Word with Squire's track changes" and email it straight to the other side.

Yes. In Word review mode Squire never regenerates your document - it amends the original file you uploaded. Styles, automatic clause numbering, headers and footers, images, and the other parties' tracked changes all remain exactly as they were. Squire's changes are added as native Word track changes, with the reasoning attached as Word comments, so opposing counsel can open the file in Word and accept or reject them as usual.

Yes. Word generates clause numbers like "3.2" or "(a)" automatically from its numbering rules rather than storing them as typed text, so Squire reads those rules and shows the exact same numbers in the review view. That means when you tell the other side "we amended clause 3.2", the number you see in Squire is the number they'll see in Word - no more guessing or recounting.

Open the redlines panel and switch to the "All changes" tab. It lists every change across the whole document - the other side's tracked changes, the AI's proposals, and your own suggestions - in document order, each labelled with its clause number and whether you accepted, rejected, or left it pending. It's the reference sheet for the note you send with the document: a clean run-down of which sections you touched and why. Click any entry to jump straight to it in the document.

Yes - select any text in the document and click "Suggest edit". Propose replacement wording or a deletion, with an optional comment explaining why. Your suggestion appears as a redline alongside the AI's proposals and the other side's changes, and exports as a track change under your own name. Direct typing is deliberately disabled so the original document stays intact for a clean export.

Click "Preview result" in the review banner. Squire shows the consolidated document - accepted and still-pending changes applied, rejected changes reverted - just like Word's "No Markup" view. Toggle back to "Show changes" to continue reviewing.

Yes. Every change to a draft is saved as a version automatically, whether it was made by you or by the AI. You can also save named snapshots at any point (for example, "Before client review" or "Signed version"), restore any earlier version at any time, and view a side-by-side comparison showing exactly what changed between two versions.

Yes. Turn on "Review AI edits" in the bar above the editor and the assistant stops editing the text directly: it proposes every change as a tracked change, and nothing reaches the document until you accept it. The Tracked changes panel opens on new proposals automatically, the bar shows how many changes are waiting, and each chat reply that proposes changes has a "Review changes" button. Accept or reject them one by one, or ask the assistant to accept them all - that is a single step you confirm. With the switch off (the default) the assistant edits directly, but every edit is saved as a version you can restore, the bar above the editor offers Undo on its latest edit (which takes back just that edit, keeping anything written since), and an edit that would delete a section still waits for your confirmation. If you are typing while the assistant changes the same paragraph, the document keeps the assistant's change there and your version is saved in Version History. Word review mode always works through proposed changes.

No. Any action that is hard to undo always requires your explicit confirmation. For example, an edit that would delete a section of your draft, accepting or rejecting all redlines, adding or removing a team member on a matter, or changing firm-wide settings will show a Confirm / Cancel prompt before anything happens - and the card says exactly what would be removed. Routine actions like making an edit you requested or saving a version happen immediately, and every edit to a draft is saved as a version you can restore.

Yes. When the assistant produces a finding worth keeping, you can ask it to save it as a note. Notes can be private (only you), shared with a matter team, or visible to your whole organisation. You can also ask the assistant to append new findings to an existing note, which is useful for building up a research log over time. Notes are searchable and appear in the matter sidebar.

Yes, for chats linked to a matter. Squire accumulates research findings across all chats on a matter, so when you start a new conversation it already has context from previous sessions. This means you don't have to re-explain the background each time, and the AI can build on earlier research rather than starting from scratch. Chats that aren't linked to a matter start fresh each time.

Yes. Just ask - for example, "set up a new matter for client X" or "add Sarah to this matter". Creating a matter happens immediately since it can always be archived or deleted later. Adding or removing a team member requires your confirmation first, since it changes who can access the matter's contents. The person you're adding must already be a member of your organisation.

Yes. A department is a matter that holds other matters - one per area of your legal work. Departments are off until an owner or admin turns them on (Organization Settings, then Matter Types, then switch on Department). Create one as a matter of the Department type, give it a team, and choose a default visibility for new matters. Open the department and choose New Matter to file a matter in it - the new matter starts with the department's team and its default visibility, which you can change afterwards. Move an existing matter in or out from its menu with Move to Department; the matter's creator or an owner or admin can do this. On the matter list each matter shows its department, and the Department filter narrows the list to one department or to matters in none. Departments sit at the top level - a department can't contain another department - and deleting a department moves its matters back to the top level rather than deleting them. More about departments.

No. A department groups matters; it never grants access. Every matter keeps its own team and visibility, and moving a matter into or out of a department changes nobody's access. The department's team and default visibility are only copied onto a matter when it is created in the department. A department itself is always visible to everyone in your organisation, so anyone can find it - but inside it they see only the matters they can open, and a private matter's title never appears to anyone but the person who made it. File sharing between a matter and its parent doesn't reach into or out of a department either.

Open the matter, open its menu and choose Pin to My Sidebar - it appears in the Pinned section at the top of your sidebar on every page. Pinning a department takes you straight to its matters. Owners and admins can choose Pin for Everyone on a matter, and under Organization Settings, then Sidebar, they can pin a whole type - such as all departments - and set the order everyone sees. Your organisation's pins come first, then your own. Hover a pin to remove it; a pin your organisation added is hidden from your sidebar only, and you can show hidden pins again at any time. Pins never give access: you only see pins to matters you can open. Pins show in the web app; the mobile app doesn't show them yet.

Click the invite button (the person with a plus) in the chat's header, or "Invite teammates" in its participants menu, and pick who to add. Anyone already in the conversation can invite - not only the person who started it. In a matter chat you can invite people on that matter's team; if the person isn't on the team, an owner, admin, or the matter lead can choose "Add to matter & invite", which also gives them access to the whole matter. You can also just mention a colleague with @ - if they aren't in the chat, Squire offers to invite them; nothing happens until you click. The people you invite see the whole conversation, including earlier messages, Squire's answers, and attached files. A chat that was just you becomes a team chat: from then on Squire answers only when someone mentions @Squire, and your personal memory is no longer used in it. Invitees get an in-app notification and an email, depending on their notification settings, and every addition and removal is recorded on the matter's activity timeline. Anyone who leaves or is removed loses access to the chat and the files attached in it straight away, and someone taken off a matter's team loses access to that matter's chats - including ones they started. A drafting chat can only be shared by the draft's owner, because inviting someone into it also shares the draft.

Yes - invite them as a guest collaborator. On a matter's People tab, click "Add collaborator", enter their email, choose what to share (the whole matter or a single document), set their access level (view only, comment, or suggest changes on review-mode drafts), and pick how long the access lasts. To bring someone into a conversation instead, open the matter chat and click "Share externally" in its header - you invite each person individually, say in what capacity (client, their counsel, counterparty, and so on), whether they can reply or only read, and when their access ends. They don't need to be in your firm: if they already use Squire it's instant, and if not they're prompted to create a free account from the invite link. Only owners, admins, or the matter lead can invite guests. The guest receives an email and an in-app notification with Accept and Decline buttons, and you're notified either way. Every invite, acceptance, and revocation is recorded on the matter's activity timeline so your team is never surprised by who has access. If an outside practice looks after your organization's legal work generally, rather than one matter, see working with a partner practice.

Open the matter, go to Documents, and click the bin icon on the document's row. You can delete a document you uploaded; owners and admins can also delete firm-wide documents. Deleting it removes it from the matter and from the AI's search straight away - the assistant stops being able to read it or cite it immediately, which is the point. The file itself is not destroyed immediately - it is retained for 30 days and then permanently deleted, along with every copy derived from it. Treat deleting as final: there is no undo in the app, so check before you confirm. Drafts work differently: deleting a draft is immediate and permanent, and it also removes that draft's chat conversation and version history, so there is no recovery window. A linked document that lives in another matter can be unlinked instead, which removes it from this matter and leaves the original untouched.

A guest sees only the specific matter, document, or chat you shared - nothing else in your firm. After accepting, they can switch into your organisation (clearly marked as a guest), where their home page is simply a list of the items shared with them. They can read a shared document and, depending on the access level you granted, leave comments or suggest changes on it. In a shared chat they see only the messages your team sends to everyone, and only from the moment they were invited - never earlier messages, your team's internal notes, or Squire's answers unless a team member chooses to pass one on. They can open files your team attaches to those messages, and with reply access they can post messages of their own. Guests cannot use Squire AI in any way, and nothing they write makes Squire respond. Access is time-boxed to the period you set and can be revoked in one click, taking effect immediately - even in a page they already have open.

Yes, you can hold a conversation with a client (or their counsel, an expert, or the other side) in a matter chat: open the chat and click "Share externally". Your client will not see what Squire says. In a shared chat Squire answers only when someone on your team mentions @Squire, and its answers are visible to your team alone. To pass one on, choose "Share with guests" on the answer - it opens in your message box so you can review and edit it, and it is sent under your name. Your team can also post internal notes the guest never sees - with the other side, an expert or anyone whose role you haven't set in the chat, your message box starts on an internal note, and you switch to "Everyone" to reach them - and anything written before a person was invited stays hidden from them. What the client tells you can be saved to the matter's knowledge with "Save to matter", recorded as their statement rather than as established fact, and you can put the matter brief's open questions to them directly and file their reply as the answer. Guests are told about new messages in the app and by email; the email never contains the message itself. Guests cannot upload files into a chat yet.

Yes. Run /extract-timelines on a court order, summons, notice of motion, or directive (or just ask "what are the deadlines in this order?") and Squire's action-timeline extractor produces a structured proposal: past procedural events go onto the matter's activity stream, and future required actions become todos with computed due dates. Deadlines expressed as relative offsets ("10 court days from service of summons") are computed using South African Uniform Rule 4 - court days exclude the first day, weekends, and public holidays. You always see a confirmation summary before any todo is created. Every todo carries a citation back to the originating page in the source document.

When you mark a predecessor todo complete, Squire automatically re-anchors and recomputes the due date of any dependent todo. For example, if "File plea (20 court days after notice of intention to defend)" depends on "File notice of intention to defend", and you complete the notice three days late, the plea's due date shifts forward to match - using the actual completion date as the new anchor. This keeps your matter timeline accurate as the case unfolds, without manually updating every dependent date.

Yes - there's a built-in trigger on file summaries that fires when a newly uploaded document mentions "court order", "summons", "notice of motion", or "directive". When it fires, Squire AI extracts the action timeline, creates the todos and activity entries, and notifies the matter team so you know automatic activity occurred. You can disable this for your organisation from /triggers - it's listed as "Auto-Extract Action Timelines" under the global triggers and can be turned off without affecting other organisations.

Yes. A workflow starts from a trigger - an event such as a document being uploaded - and then runs its steps on its own. Open Workflow -> New Workflow (or go to /workflows) and choose how to start: pick a ready-made template, start from scratch in the builder, or ask the Squire team to build one with you. The builder reads as a sentence - When [an event happens] · Only when [optional filters] · Then [the steps Squire runs] - and narrates back in plain English what it will do. Events include a new matter opened, a document uploaded, a task completed, contract milestones, table changes, or a button you click on a matter. Steps can ask Squire AI to analyse or draft, create to-dos or notes, notify people, pause to ask someone for input, update records, or drive a table. "Save & test" runs the workflow against sample data so you see the result before it goes live, and AI steps pass an automated safety check. Templates include new-matter intake, summarising new documents, a one-click matter status report, and a human-in-the-loop document review. More about workflows.

Yes. Add a "Request User Input" step (under "Involve a person" in the workflow builder) and the workflow stops at that point and asks a person - usually whoever started it, for example by uploading the document - to approve the work or add comments. They get a notification, and the request waits under Workflow -> Pending Actions until it is answered; a request can expire if it is not answered in time. Only after it is answered do the later steps run, such as saving a note or telling the team. The "Human-in-the-loop document review" template is set up this way: Squire drafts a review of each new document, and the review is saved only after a person has seen it. Owners and admins can see every run, step by step, under Workflow -> History.

Squire has a built-in Features page at /features - a guided tour of the product, with a card for each capability and a detail page explaining what it does and where to find it. It covers document drafting, Word review mode, Tables, action-timeline and deadline extraction, notes and matter memory, entity and conflict detection, skills and commands, workflows, guest collaboration, client conversations, and more. Newly shipped features appear there first with a "New" badge, and you'll see a short one-time announcement on your dashboard the next time you log in. Open it any time from the Features card on the Support page, or go straight to /features.

Yes. You can connect an AI assistant you already use - Claude, ChatGPT or Claude Code - to your records in Squire, and ask it things like "what's overdue across my matters?" or "has the supply agreement been signed?". In Claude, add a custom connector (Settings, then Connectors); in ChatGPT, add a custom connector; in Claude Code, run claude mcp add --transport http squire with the link and then /mcp. Paste Squire's MCP link - shown on the assistants feature page - sign in to Squire, and choose what the assistant may do. It starts read-only; you can let it add and update tasks, and limit it to particular matters. It can read your matters and their recent activity, your tasks and deadlines across matters, notes, file names and summaries (not the files themselves), contracts with their key dates, approval and e-signature status, drafts, tables and templates. The connection appears under Settings, API Keys as a connected app, and you can disconnect it there at any time.

Yes. Squire has a REST API - for matters, tasks, notes, files and chats - and an MCP server, which is how AI assistants and your own agents connect, with tools for tasks and deadlines, matter activity, contracts, drafts, tables and templates. Both use API keys you create under Settings, API Keys: each key holds only the permissions you give it (there's a read-only preset), can be limited to particular matters, and can be revoked at any time. Assistants and agents connected over MCP work with your records only; they never use Squire's own AI. The reference is at /docs/api (sign-in required); to connect an assistant such as Claude or ChatGPT you don't need a key at all - see the assistants feature page.

Table views are a spreadsheet-style surface where each row is an entity (a file, matter, todo, contact, etc.) and each column is an AI-extracted answer or summary. They're built for the comparison work that doesn't fit in a chat: "across these 30 NDAs, what's the assignment clause?", "for every open matter, what's the next deadline?", "list every active client with their billing contact". You can create one from /tables or ask the assistant to suggest one mid-conversation.

A table's rows come from a row source you pick when creating it. The supported sources are: files (your org file library), matters (your matter list), todos (tasks across matters), team members (people on a matter), notes, entities (people or companies tracked on a matter), clients (your client list), and custom (you type the row labels). Each source carries the right context to the AI - file rows let the AI cite specific pages of the source document; matter rows let it reference linked files and metadata.

Each column has a name and a prompt (e.g. "Termination notice period - answer in days, or 'N/A'"). When you click Regenerate, Squire runs the prompt against every row in parallel - locked cells are skipped - and writes the result with a citation back to the source page where applicable. Click any cell to open the inspector and see the AI's reasoning and any citations. Every cell is versioned, and you can edit any cell inline to override the AI.

Yes. Once a table has at least one populated cell, the header shows a download button that exports the table as an Excel workbook (.xlsx) with three sheets: the table itself, the citations, and metadata. New tables can also start from a template: blank, or one of the seeded system templates - Document review, SPA diligence, Open todos triage, Matter portfolio dashboard, Custom comparison. Templates carry a row source and pre-built columns so you can populate the table in one click instead of designing it from scratch.

Yes. Each table has its own chat pane on the right. You can ask the assistant to add or update columns, regenerate specific cells, set values, lock cells against AI overwrite, or refresh the row source - destructive changes ask for confirmation first. Tables also integrate with workflows: triggers can fire on table_view.created / .row_added / .cell_changed / .generation_finished events, and four trigger actions (populate_table_view, add_table_row, update_table_cell, regenerate_table_cell) let workflows drive table generation and writes.

Squire tracks contracts as first-class records, not just files. Every contract at /contracts has a lifecycle (draft → negotiation → approval → executed → active → expired/terminated/renewed), key dates that drive reminders, parties, attached documents, and an approval trail. You can create a contract from a Squire draft or import existing signed contracts, then filter the repository by status, type, counterparty, or expiry window. It's one system of record from first draft to renewal notice, with the AI already inside it.

Use the Import page at /contracts/import - drop up to 20 executed contracts at a time. Each file becomes a contract record with the document attached, and Squire AI reads it to extract the parties, key dates, value, governing law, and renewal terms, with the exact source quote for every value. Extracted contracts land in a review queue where you confirm (or correct) the values before they count. Note: scanned documents without a text layer can't be read yet - they're flagged honestly in the review queue so you can fill the fields in manually.

You verify them - that's the design. Every extracted date and value is shown next to the exact contract wording it came from, with a confidence score. Until a person confirms a contract's extracted metadata in the review queue, those dates are treated as unconfirmed: they never create reminders, never fire automations, and Squire AI warns about them in chat. A hallucinated expiry date causing a missed renewal is exactly the failure a contract system exists to prevent, so the human confirmation step can't be skipped.

From a contract's Approvals tab, submit it to an ordered chain of approvers from your organisation. Approvals are strictly sequential - each approver is notified in-app and by email when it's their turn, and only the current step can be decided. Rejection returns the contract to negotiation with the reason recorded; when the last approver signs off, the contract moves to approved and execution can be recorded. Everyone's queue lives at /approvals ("My Approvals" in the sidebar), and every decision is recorded on the contract's activity timeline.

Key dates on an active contract - the expiry date and the renewal notice deadline - become todos on the contract's matter, and Squire sends reminders as they approach: in-app and by email at 14, 7, and 1 day out. Auto-renewing contracts roll their expiry forward automatically at the end of each term, and approaching expiry can also fire your own workflow automations via the trigger engine (contract.expiring). Every reminder email includes one-click unsubscribe, and you can switch to in-app-only from notification settings.

Yes - Squire gathers signatures itself, no DocuSign account needed. From a contract's Documents tab choose "Send for signature": pick the PDF, add signers (name, email, signing order), and drag the signature, date, and text boxes onto the page. Signers need no Squire account - they get an emailed link, confirm a one-time code sent to their inbox, agree to sign electronically, and sign by typing or drawing their name. Signing runs in order, and you can nudge or withdraw at any point. When everyone has signed, Squire seals the document cryptographically, attaches the signed copy and a completion certificate to the contract, and moves it to executed. You can still record execution by hand for wet-ink and counterparty-papered deals. Because signing happens in Squire, the document never leaves your region - it is not sent to a third-party e-signature provider.

Electronic-signature law generally recognises tiers, and Squire produces the base one - a simple electronic signature (also called basic or ordinary). It is recognised under the electronic-transactions law of the jurisdictions Squire operates in - ESIGN and UETA in the United States, eIDAS in the EU, ECTA in South Africa, and their equivalents elsewhere - and is sufficient for most commercial contracts where no statute requires more. What makes it defensible is the evidence: who signed, the one-time code they passed, the exact wording they consented to, the SHA-256 of the document they were shown, and a tamper-evident log of every step - all set out in a completion certificate attached to the contract. Squire also seals the finished document so any later change to it is detectable, and anyone can check a signed PDF at /verify without an account (the file is fingerprinted in your browser and never uploaded). Squire does NOT produce an advanced or qualified electronic signature (the eIDAS terms; an accredited advanced signature in South Africa) - those are regulatory accreditations held by licensed providers, not product features. Every jurisdiction also carves out documents that can't be signed this way: wills and testamentary documents almost everywhere, and depending on where you are, land and property transactions, certain guarantees or negotiable instruments, and some family-law or court documents. The specific list is a matter of local law - Squire shows you the exclusions that apply to your deployment and asks you to confirm your document isn't one of them before it goes out. Where a statute requires a particular form of signature, use an accredited provider.

Yes. In any chat, the assistant can list and read your contract records - lifecycle status, key dates and the renewal notice deadline, parties, documents, approval state, and the extracted summary - always limited to contracts you can access. For portfolio-wide questions ("what's the termination clause across all our supplier agreements?"), start from the Contract repository table template, which extracts answers across every contract with citations and exports to Excel.

Because it is the evidence. Once a document has been sent for signature, that exact file is what everyone signed against, and the signing record refers back to it - so Squire keeps it permanently and the delete button is disabled with the reason shown. If you no longer want the signature request to go ahead, void it instead: that takes the request out of circulation and stops anyone signing, while leaving the record of what happened intact. The same applies to the current executed copy attached to a contract - remove it from the contract's Documents tab first if you need to replace it, and to documents attached to a support conversation.

You choose per contract: organisation-wide (any member), following the matter (a contract created inside a matter inherits that matter's visibility by default, so team-scoped or private matters keep their contracts confidential), or private to you. Approvers must be able to see a contract to be added to its approval chain, and the AI sees exactly what the asking user could see.

A guest is for one piece of work: you invite them to one matter, document or chat, they can view, comment or suggest changes, they can't use Squire's AI, and their access ends when you choose. A partner practice is a legal practice - often fractional in-house counsel - that looks after your organization's legal work more generally: once one of your owners accepts its proposal, its lawyers work across your organization as lawyers do, with Squire's AI, and everything they open is recorded for you. For an advocate briefed on one case, or the other side's lawyer, use a guest (sharing with someone outside your firm). For the practice that is, in effect, your legal team, use a partner practice (Counsel from a partner practice).

Yes, if the practice is a Squire partner. Squire proposes the engagement to your organization: your owners are emailed, owners and admins see the proposal and its terms on Organization → Counsel, and only an owner can accept it. Accepting is your consent, recorded with who accepted, when and on which terms; until then the practice's lawyers can open nothing. Once it's accepted, its lawyers appear in your organization named with their practice ("Jane Doe · Acme Counsel"). If your practice isn't a Squire partner yet, ask them to contact us.

They work as lawyers in your organization, never as owners or admins. They see the matters open to your organization and those they're added to, never a colleague's private matters or chats, and they can't change your settings, members or billing; your owners and admins can let them set up your matter types and the sidebar pins everyone sees. Your owners and admins see Counsel activity: every page counsel open, action they take and file they download, person by person, with monthly figures. If the terms you accepted say so, the practice sees those monthly figures too (days active, actions and AI calls), never what its counsel worked on. Your members can send counsel a request from the counsel panel in the sidebar; only the person asking and counsel see it.

Counsel take none of your seats. Who pays for your own seats depends on your arrangement with the practice, and differs from one engagement to another: you may subscribe to Squire yourselves as usual, or the practice may provide Squire as part of its service - reselling it to you - and pay for your seats. Your billing page says which, and when the practice pays, it can set how many seats you have. If a practice stops paying for you, you get 30 days of full access to choose a plan of your own. See pricing for subscribing directly.

On Organization → Counsel, an owner or admin can Suspend the engagement, which pauses counsel's access and removes nothing (and resume it later), or End engagement, which removes their access for good. What counsel created stays in your organization. If the practice was paying for your seats, ending the engagement gives you 30 days of full access to choose a plan.

A partner practice can offer the organizations it serves its own templates, table templates, skills, house style and matter setup (matter types, sub-types and pins), one at a time or bundled in a pack. Your owners and admins install copies from Their library, in your counsel panel; counsel you've allowed to set up your organization can install templates and the matter setup. A copy is yours to change. When the practice publishes a newer version you see what changed - and what you'd edited in your copy - and choose to take it, keep yours, or install it alongside. A matter setup only adds: it never removes anything of yours, and a matter type you've turned off stays off. A practice's owners can also put a pack's templates straight into your organization; the rest of the pack waits for your owners or admins.

Yes, as a Squire partner. Your lawyers work inside each client's own Squire organization once its owner accepts, each client in its own browser tab, named with your practice. Your Clients page lists every client you serve, with each person's unread notifications and open tasks there. You can offer clients your templates, skills and house style through your library, put your logo, colour and wording where they meet you, and - depending on the arrangement - provide Squire to them and pay for their seats on one monthly invoice. Squire sets partners up: see Squire for counsel practices and talk to us from there.

A connected assistant sees only what you can see in Squire, and only what you allowed when you connected it: it starts read-only, and if you limit it to particular matters it reaches nothing outside them. It works with your records only - it cannot ask Squire's own AI anything, and Squire's AI does not reply to messages it posts. It cannot approve, sign, or edit drafts; those stay in Squire with a person. Tasks it adds or completes show as yours, made through the assistant. If your organization requires two-step sign-in, a connection needs it too, and leaving the organization disconnects every one. Owners and admins can turn connected assistants - or all API access - off for the whole organization in Organization Settings, Security; that pauses existing connections until it is turned back on. One thing to weigh before connecting: whatever the assistant reads leaves Squire and is handled under your own agreement with that assistant's provider, not under Squire's no-training terms or its pseudonymisation - so connect only assistants your firm permits for client information.

Your firm's data is stored in your jurisdiction, and your uploaded documents, AI chats, drafts, notes, the descriptions of your matters and tasks, the parties and clients on your matters, your tables and what your workflows produce are encrypted at rest under a key unique to your firm. The AI analysis is done by enterprise AI providers working from pseudonymised text, and today that analysis may run outside your jurisdiction; we are moving it onto servers we operate in your jurisdiction, starting with personal-information detection. On-premise deployment is available for Enterprise customers by arrangement - custom integration work that depends on your network and hardware, scoped and priced with you.

Squire combines per-firm envelope encryption, pseudonymisation, and strict access controls. Each firm gets its own encryption key (AES-256-GCM), and the content of its work is encrypted under it, each piece sealed to its own record: uploaded documents and their summaries; AI chats and the research behind them; drafts, redlines and comments; notes; the descriptions of matters and tasks; the parties and clients on its matters, with their contact details, and its conflict checks; its tables - their rows, cells, the AI's reasoning and citations, and each cell's history; and what its workflows produce. Titles, file names, table column names and the topic words Squire uses to find your work stay readable, so search keeps working. Your members' names and email addresses stay readable too: sign-in needs them. Squire still finds a party by name, matching inside the app rather than in a readable index. Before a document reaches the AI that analyses it, the people in it are replaced with reversible pseudonyms - a separate, single-purpose model finds the names, and the mapping stays in the firm's encrypted catalog. Access is governed by roles and per-matter permissions. Every account can add two-step sign-in - a passkey or an authenticator app - and a firm can require it of everyone who works in it.

No. Squire does not train AI models on your data, and your data is only accessible to authorised users in your organisation. Squire does not train AI models on anyone else's data either: its answers are grounded in curated legal sources and in your own documents, which it reads to answer your questions rather than learns from. The AI providers it uses are contractually barred from training on your data.

Squire is designed to meet POPIA and GDPR requirements - South Africa's Protection of Personal Information Act and the EU's General Data Protection Regulation. Your firm's data is stored in your jurisdiction, its documents, chats, drafts and notes - and the parties and clients on its matters, its tables and what its workflows produce - are encrypted under a key unique to your firm, and documents are pseudonymised before AI analysis - pseudonymisation and encryption being the two security measures the GDPR names. Beyond that, Squire can trace where a person's details appear in your firm's records to answer an access request, supports erasure requests, and logs its AI requests as proof of what was sent without keeping the content. That is the operational evidence a responsible party needs to answer a data subject or the Information Regulator, not just a compliance claim in a privacy policy.

No - the assistant loses access the moment you delete it. Deleting a document immediately removes the searchable text the AI reads, so it can no longer find, quote, or cite that document in any answer, and any background processing still running on it is cancelled. Copies held by the AI providers we use are deleted too. The document leaves every list, search result and picker at the same time. The file itself is retained for 30 days and then permanently destroyed along with every derived copy - the extracted text, the de-identified text and the search index. There is no undo in the app, so treat deleting as final. Deleting a draft is immediate and permanent, with no recovery window. Some documents cannot be deleted at all, because they are signing or contract evidence; Squire tells you which and why rather than failing silently.

No. Guest access is deny-by-default: a guest is never a member of your organisation and can only reach the exact matter, document, or chat shared with them. Sharing a matter does not share its chats - each conversation is shared separately, person by person, so a lawyer's own research chats on the matter stay private. They never see your other matters, your firm-wide content, or anything not explicitly shared, and they cannot interact with Squire AI. Every grant is time-boxed with an expiry you choose and can be revoked at any time, with access ending immediately. All guest invites, acceptances, declines, and revocations are logged on the matter's activity timeline, so your team always knows what's shared and with whom.

Yes. In Settings → Security, choose Set up an authenticator app, scan the QR code (or paste the setup key) with the app you already use - the Passwords app on iPhone, iPad and Mac, Google Authenticator, Microsoft Authenticator, 1Password or Authy all work - and enter the six-digit code it shows. From then on, signing in with your password or an emailed link also asks for a code from that app. A passkey sign-in counts as both steps on its own, because the passkey already checks your fingerprint, face or PIN. When you turn it on you get ten single-use recovery codes for a lost phone; keep them somewhere safe. Squire does not send codes by text message. If you lose both your phone and your recovery codes, Squire support can reset two-step sign-in, but only after confirming who you are with your firm.

Yes. An owner or admin opens Organization Settings → Security, chooses the date it starts - two weeks out by default, at most 90 days - and turns it on. Members who haven't set up a passkey or an authenticator app yet are emailed, and nothing is blocked until that date. From then on, a password or an emailed link alone no longer opens the firm's work - in the browser, the API, or AI tools connected to Squire: people who have a passkey or an app confirm it's them, and people who have neither are asked to set one up first. It covers members, counsel working for the firm, and guests the firm has shared work with. No one is removed from the firm, and on the Team tab owners and admins can see which members and counsel sign in with a passkey, an authenticator app, or a password only. Enterprise firms can also sign in through their own identity provider with single sign-on, which is set up by arrangement.

Squire runs as a managed service, with your firm's work encrypted under a key unique to your firm and stored in your jurisdiction. On-premise deployment is available for Enterprise customers by arrangement. It takes custom integration work that depends on your network and hardware, so it is scoped and priced with you - talk to our team about what your firm needs.

"Hallucinations" occur when generative AI makes up false information. In a legal context, this can lead to dangerous misinformation. Users must be wary because even plausible-sounding answers can be completely incorrect. Squire addresses this risk with its Hallucination Firewall, which grounds answers in legal sources and checks citations against them - though you should still verify anything you rely on.

Squire's hallucination firewall grounds answers in actual legal text: every claim is linked to a source, citations are verified against the original document, and each response is scored for how well it's supported. Instead of inventing authorities, Squire surfaces the underlying sources so you can confirm them. This minimises AI errors and protects your firm's reputation - though a qualified professional should still review anything material.

Squire provides research assistance, not legal advice. Its answers are grounded in official legal sources and local commentary, and every answer comes with references so you can verify them against the source. Check the sources before relying on an answer, and have an attorney review AI suggestions before final use.

Yes. Squire's database is regularly updated. This ensures Squire's answers reflect current statutes and precedents in your jurisdiction.

Squire is version-aware. When a statute or regulation has several versions in the corpus - earlier and later amendments, consolidations, and published drafts - Squire ranks the current in-force version first and labels each result ("In force", "In force (later amendments exist)", "DRAFT - not in force", or "Repealed"). It pins a single authoritative consolidated version per topic and won't present a repealed or draft provision as operative law - it flags it instead. For comparisons it keeps both items within the same consolidated version rather than mixing an old sub-section with a newer one.

Squire generates its own analysis, not published opinions, so it's treated as a research tool. You can cite the original sources it provides (cases, statutes). We recommend citing the primary authority rather than Squire's text. Think of Squire like a very fast research assistant that points to what to cite.

Squire has three plans: Student (coming soon), Professional, and Enterprise. Student and Professional are billed per seat at a fixed monthly fee, with no token or usage-based charges; Enterprise is custom-priced. There is no free plan - Where Do I Stand? is our free public tool. For current prices and what each plan includes, see the pricing page.

Yes, by approval. Access to the full platform is by approval: book a demo with our team, or join the waitlist. Approved organisations get a trial with no credit card required - see How does the trial period work? for what's included. Where Do I Stand? is our free public tool, with no account or payment needed.

We use fixed pricing for predictability. You pay a set monthly fee per user (like any software subscription). No hidden cloud or token fees means firms can use Squire freely without watching usage meters. This model suits legal budgets and lets you focus on value.

The Enterprise plan adds premium support and priority access to new capabilities. On-premise deployment is available for Enterprise customers by arrangement, as is single sign-on with your identity provider - both are custom integration work that depends on your systems, network and hardware, scoped and priced with you. It's ideal for large firms. Please speak with our sales team to tailor a package to your needs.

Pricing is per seat. Each active member of your organisation is one seat, billed monthly or yearly. The Subscribe flow asks you to pick a seat count, and you can adjust it later from the billing page or through Stripe's Customer Portal. Your total is simply your per-seat price multiplied by the number of seats - see the pricing page for current per-seat rates.

Yes, if your plan has yearly pricing configured. At subscribe time, you can toggle between Monthly and Yearly on the plan picker. Once subscribed, the current-plan card offers a "Switch to yearly" (or "Switch to monthly") button when you toggle to the other cycle. Stripe prorates the difference on your next invoice - so switching mid-cycle is safe and immediate.

Access to Squire is by approval. When Squire approves and invites a new organisation (from the waitlist or after a demo), the organisation starts a free trial with full access and no credit card required. Joining an organisation that already uses Squire starts no trial of your own: you share that organisation's access, trial or subscription. A trial's end date is set when the organisation is invited - usually the end of the following month - and the invitation email to its point person and the organisation's settings page give it. We send reminder emails 7, 3, and 1 day before it expires. When the trial ends, the organisation enters a 14-day grace period: most features become read-only, but the billing page stays open so you can subscribe. If the grace period passes without a subscription, access is suspended.

On /organization/billing, click "Add seat" on the Subscription card. This calls Stripe to increase your seat count by one, and Stripe prorates the cost on your next invoice. You can also adjust seats through Stripe's Customer Portal (click "Manage billing"). We use pre-paid seats: you can invite up to your purchased seat count, and beyond that the invite flow will prompt you to add a seat first.

The invite is blocked with a message explaining you've reached your seat limit, and you're redirected to the billing page. From there you can add another seat in one click - Stripe will prorate the cost - and then return to invite the new member. This keeps billing predictable: you never get surprise charges just because someone accepted an invite.

The billing page at /organization/billing shows your invoice history with links to download each one (hosted by Stripe or as PDF). To update your card or change your payment method, click "Manage billing" - this opens Stripe's Customer Portal where you can also cancel the subscription or change billing details. All invoice emails come from Stripe with full VAT breakdown where applicable.

Yes, you can change plans at any time. Pricing will be pro-rated by day.

Where Do I Stand? is a free AI legal analysis tool on Squire.law that lets anyone describe a legal situation and receive an instant, structured analysis based on South African legislation and case law. It covers areas including employment law, contract law, family law, personal injury, business law, criminal law, real estate, immigration, intellectual property, and constitutional law. No account or payment is required.

Yes, completely. You don't need to create an account, join a waitlist, or enter any payment details. Describe your situation, select your jurisdiction and legal area, and receive your analysis within seconds. It's designed to give you a useful starting point - and to show you what Squire's full platform can do for legal professionals.

The tool currently supports analysis across contract law, real estate law, employment law, family law, personal injury, business law, immigration law, criminal law, intellectual property, and constitutional law. You can also select "Other" for legal situations that don't fit neatly into these categories.

You can select from South Africa, Namibia, the Netherlands, the United Arab Emirates, or General International. The AI's analysis is tailored to the legal framework of whichever jurisdiction you choose - so a South African employment law query will reference the Basic Conditions of Employment Act and Labour Relations Act, not US or UK legislation. Note that this wider choice is for Where Do I Stand? only: the full Squire platform is currently live in South Africa only.

No. The tool provides general legal information and analysis - not legal advice. It's designed to help you understand your situation, identify relevant legal issues, and see what the law says. For serious legal matters, you should always consult with a qualified attorney or advocate who can assess the full facts of your case. Think of it as an informed starting point, not a replacement for professional legal counsel.

The tool uses the same AI engine that powers Squire's full platform - grounded in verified South African legislation, case law, and legal commentary, with a hallucination firewall that validates citations. It aims to provide well-grounded legal information with relevant legal authority. However, legal situations are fact-specific, and AI cannot account for every nuance. Always verify important findings with a qualified legal professional.

The more specific you are, the more useful the analysis. Include relevant dates, amounts, locations, the parties involved, what has happened so far, and what outcome you're hoping for.

Squire takes on research and analysis that traditionally take hours: it searches statutes and case law, highlights key clauses, and drafts summaries for you to review. Lawyers spend less time on routine searching and first drafts, and more on strategy and judgement.

Squire excels at document-heavy tasks: contract review, due diligence, and case preparation. For example, it can compare agreements and extract obligations. Use Squire to analyze discovery documents, generate first drafts, or check compliance. Essentially, any workflow involving large legal texts can be accelerated by Squire.

Mid-size and large law firms and corporate legal departments see the greatest return. Professionals in litigation, corporate law, and regulatory law gain rapid insights from Squire's analytics, and a Student plan for law students is coming soon. In short, any legal team aiming for data-driven efficiency will benefit from using Squire.

Calculate savings by comparing hours spent on research/drafting before and after Squire. For example, if Squire cuts research time from 5 hours to 30 minutes per matter, that's a 90% time saving. Factor in billable rates to quantify revenue impact. We also provide analytics tools to track time saved and efficiency gains.

The full Squire platform is currently live in South Africa only. Its answers are grounded in South African constitutional, statutory, and case law. Where Do I Stand?, our free public tool, also supports Namibia, the Netherlands, the UAE and General International. We have plans to expand the full platform into African and Middle Eastern markets; Saudi Arabia and others are coming soon.

Yes. We are actively adding new jurisdictions. Each new jurisdiction gets its own curated legal sources. If you have a specific country in mind, please contact us to discuss partnership or early access.

Currently, Squire's interface and legal content are English-based.

Squire's legal sources include both common law and statutory (civil) sources as applicable. For example, its South African sources cover the country's mixed legal system, including Roman-Dutch law. Squire does not train a model per country: for each jurisdiction, its answers are grounded in sources curated for that country's system, whether common law, civil law, or hybrid.

Yes. Since Squire's database includes extensive statutory and case law, it can answer questions in any practice area covered by those sources. If your firm has specialized needs (e.g. local IP statutes), you can upload those documents to Squire's private data layer to extend its expertise.

No. Squire is a research assistant, not a lawyer. Its answers are for informational purposes only. We recommend having a licensed attorney review any AI output before acting on it. Squire's purpose is to augment your expertise, not replace professional judgment.

Squire grounds its answers in authoritative legal sources, which reduces bias. We also vet content for offensive or unethical language. Because its answers draw on professional legal sources it reads and cites rather than on the open web, it avoids many common bias issues. However, users should remain vigilant - the underlying AI models are not free from all bias, so we encourage human oversight.

Laws regulating AI are emerging, but currently no major jurisdiction bans legal AI tools. However, you should comply with attorney ethics rules (e.g. confidential data must remain protected). Squire's design (data stored in your jurisdiction, per-firm encryption, and pseudonymisation before AI analysis) helps you stay within data privacy laws. We also advise consulting your legal ethics board if unsure about AI use in your firm.

All paid plans include in-app support chat, email support, and access to online documentation. The in-app chat is powered by Squire AI, which can answer product, account, and billing questions instantly. If you need a human, you can invite a Squire support staff member into the chat with one click. Our team is available to help with setup, onboarding, training, and troubleshooting.

Yes. Every account has a built-in product support chat at /support/chat. Squire AI answers product, account, billing, and usage questions instantly, using context about your organization and activity. If the AI can't help, you can invite a human Squire support staff member into the chat with one click - they only see the conversation after you tick an explicit consent checkbox. Support chat is for product help only; for legal research use your regular Squire chat.

Yes. Support chats have a side panel where you can upload screenshots, PDFs, and short log/text files (10 MB max, 5 files at a time). These attachments are intentionally kept out of the chat message stream - they are visible only to human support staff, and Squire AI cannot see or interpret them. If you want the AI to help with the contents of a file, paste the relevant text into the chat instead. If you need a human to look at the file, click "Invite support staff" and a Squire support agent will review it after you confirm.

For all plans, we include initial onboarding training for your team. Training covers system usage and best practices. Our goal is that your firm is fully comfortable with Squire's capabilities. Training can be booked here.

You can submit bugs or feature requests through our support portal or by emailing contact@squire.law. Feature requests from multiple firms are regularly reviewed for inclusion in our roadmap.