Frequent questions.

SafeIdea is compliance infrastructure for AI in legal practice. It is a local application that runs on the attorney's machine and makes any AI usable on confidentiality-sensitive work. SafeIdea's patent-pending Masking Engine identifies sensitive entities in your documents and replaces them with stable masked placeholders before anything is transmitted to Claude, ChatGPT, Gemini, or your legal-specific platform. The original never leaves your control. Firmwide masking rules apply uniformly to every prompt across every attorney, so the firm's approach to confidentiality is consistent rather than improvised. You can produce a Compliance Receipt for any matter on demand, a tamper-evident artifact that documents what action was taken, when, and under which firm masking rules, without ever capturing the underlying client content. The interface is built for the way legal professionals actually work: drop a document onto SafeIdea, name the matter, and get to work.

No. SafeIdea works with Claude, ChatGPT, Gemini, and legal-specific platforms like Harvey, CoCounsel, Vincent, Legora, and Irys. You keep your existing subscription, your existing workflow, and your existing AI relationships. SafeIdea sits between you and whatever AI you choose. It is the compliance layer, not a competing application.

Contractual protection and architectural protection are not the same thing. A no-training clause depends on the vendor's policies holding up under subpoena, breach, or a unilateral terms change. Heppner and Wang are both recent decisions that have unsettled the assumption that AI conversations stay private: Heppner held that a defendant's own AI conversations were not protected by privilege on the facts of that case, and Wang held that AI chat logs are discoverable from the provider. SafeIdea answers Rule 1.6 with architecture rather than with someone else's contract. Confidential client information never leaves your machine. There is nothing to subpoena.

No. We claim something narrower and more useful. A documented local-first workflow, with the patent-pending Masking Engine, attorney-approved review, and an auditable Compliance Receipt for every matter, is a documented "reasonable efforts" workflow under Rule 1.6 and ABA Formal Opinion 512. The standard is not absolute confidentiality. The standard is reasonable efforts, and reasonable efforts now has a documented form.

You see every proposed mask before anything is transmitted. The Masking Engine identifies candidate entities, presents them to you for review, and you approve, edit, or add to the list. The attorney is the expert in the room. SafeIdea makes that expertise executable rather than replacing it. Nothing leaves your machine until you say so.

You could. Many careful attorneys do today. They mask sensitive information, run the cleaned version through the AI, and reassemble the result by hand. It is a productivity sink that defeats the purpose of using AI. SafeIdea collapses that workflow into a few clicks while keeping you in control of every masking decision. The reassembly happens on your machine, not in your head.

The opposite, in practice. When you decide what the AI sees, you control output reliability. Names, email addresses, and party identifiers carry confidentiality weight. Contract amounts, jurisdiction, and clause structure rarely do. SafeIdea makes that distinction operative. You give the AI what it needs to be useful, and nothing more. Output gets better, not worse, because the AI is working on the right information.

They architecturally cannot. The cloud is where their product lives. The whole value of SafeIdea is that confidential client information never reaches a cloud. A frontier AI vendor offering "local compliance" from their cloud is a contradiction in terms. SafeIdea exists because the legal profession needs a layer the AI vendors cannot build.

The profession has known the principle for a long time. Protecting confidential client information ultimately depends on keeping it under the attorney's physical control. Rule 1.6 has implied this since 2012; ABA Formal Opinion 512 made it explicit.

What has changed is not the principle but the architecture of the tools attorneys are adopting. Frontier AI requires data to deliver reasoning. The duty of confidentiality requires that client data not be shared with third parties the attorney does not control. The conflict between those two facts cannot be resolved by contract or by terms of service. It is structural.

Heppner and Wang are recent evidence of where the conflict surfaces. Heppner found that on its facts (a defendant's own consumer-AI use, without attorney direction and under terms that permitted disclosure to the provider) there was no reasonable expectation of confidentiality and no privilege protection. Wang ordered an AI provider to produce user chat logs in discovery. These rulings may be revised, distinguished, or overturned; the structural problem they reveal will not be. The practical question is no longer whether to use AI, but whether you can show a regulator, a client, or a malpractice carrier that confidential information stayed under your control. SafeIdea is the documented answer to that question.

A per-matter artifact you produce on demand, that documents who used SafeIdea, when, and what kind of action was taken on the matter. The Compliance Receipt is signed and cryptographically chained so it is tamper-evident, and it never captures the underlying confidential content. It is the artifact you hand to a managing partner, an ethics counsel, a malpractice carrier, or a court if asked. It is what makes "reasonable efforts" tangible rather than rhetorical.

Yes. SafeIdea installs on your machine, works with the personal AI subscription you already pay for, and lets you produce a Compliance Receipt for any matter on demand. As a solo, you are also your own Firm Administrator. The role exists in the product whether you are one attorney or twenty, and you choose how much firm-level structure to set up. There is no IT department to coordinate with, no enterprise procurement cycle, and no minimum seat count. Drop a document onto SafeIdea, complete a first masking event in about five minutes, and you have a defensible workflow. If you later add an associate or staff, you scale up by purchasing additional seats at the same per-seat rate. One product, one price per seat, whether you are one attorney or twenty.

Original documents never leave the workstation. SafeIdea's cloud services handle authentication, license verification, and usage metering, none of which touch client content. Chat history stores locally. The architecture is built to make a security review short rather than long. Your IT team can confirm the data path in a single conversation.

Access to SafeIdea requires authenticated credentials. Each license is a seat managed by your Firm Administrator, who assigns the email address that uses each seat and can reassign as needed. Reassignment automatically terminates the previous email's access. One seat supports the assigned attorney's multiple personal computers (for example, office machine and home machine); each install is a data island. SafeIdea's cloud services for authentication, license verification, and usage metering use industry-standard encryption in transit. Because client documents, prompts, and chat history never reach SafeIdea servers, the most consequential attack surface in cloud-AI workflows, the breach that exposes confidential client content, does not exist in the SafeIdea architecture by design. Additional security detail will be added to this answer as the SafeIdea security program matures.

Your firm sees the firmwide masking rules it has configured (its Firm Masking Dictionary). The firm does not see the contents of your prompts, the substance of your matters, your individual Compliance Receipts, or any per-attorney usage statistics. Compliance Receipts are your artifact; you produce one when you need to show your work to a managing partner, ethics counsel, a malpractice carrier, a regulator, a court, or a client. The confidentiality boundary holds inside the firm by design.

Your Firm Administrator builds the Firm Masking Dictionary once with the SafeIdea Indexer, then places the resulting file in any folder your firm's SafeIdea installs can read from disk. Each attorney's SafeIdea then points to that folder via Settings › Indexes, a one-time setup per attorney.

In practice, the easiest place to put the file is a folder that's already synced or mounted to every attorney's machine, a OneDrive or SharePoint sync folder, a Dropbox folder, or a mounted network share. The Firm Masking Dictionary stays inside the storage your firm already trusts and audits, under the same controls you use for matter documents. SafeIdea reads from a local path; whatever cloud or on-premise system that path resolves to is your choice.

SafeIdea does not include its own sync service, by design. When your Firm Administrator refreshes the dictionary against new seed sources, attorneys pick up the new version the next time their SafeIdea reads from the shared folder.

If you are a solo practitioner, the dictionary just lives on your own machine, no sharing setup is required.

SafeIdea is $399 per seat per month at list, billed monthly, with $359 per seat per month (10% off, $4,308 per year) on annual prepay. One SKU, one per-seat price, no package distinctions. Self-serve checkout up to 199 seats; contact us for larger deployments.

SafeIdea is currently in the Charter Customer program. Charter Customer pricing is $249 per seat per month, locked for your first 12 months from first billing. See question 16 for how the Charter Customer program works.

A 30-day money-back guarantee applies to all purchases. SafeIdea does not mark up inference. Your AI subscription costs sit where they already sit, on your existing relationship with Claude, ChatGPT, Gemini, or your legal-specific platform. There is no AI vendor lock-in inside the SafeIdea price.

The Charter Customer program is how SafeIdea is launching. It runs in two phases.

During the closed testing phase, SafeIdea is admitting Charter Customers on a reviewed basis. Apply through the pricing page. We review applications and admit attorneys at our discretion. Admission triggers a payment link; we do not request your card until we have admitted you. Admitted Charter Customers pay $249 per seat per month for the first 12 months from first billing, then convert to standard pricing ($399 monthly or $359 with annual prepay) on the 12-month anniversary of first billing. Applicants who apply during this phase but are not admitted immediately are still part of the Charter Customer cohort; they receive admission when the program opens to all.

When the program opens to all, self-serve checkout becomes available, the Charter Customer price of $249 per seat per month remains the same with the same 12-month lock from first billing, and a 60-day window applies. After 60 days, the Charter Customer program closes and standard $399 pricing applies to all new signups.

Annual prepay during the Charter Customer window is $2,988 per seat per year ($249 x 12), flat.

After admission, same day. Download, install, drop a document onto the workspace, name the matter, and your first Matter-Level Masking Dictionary is created locally. The first masking event takes about five minutes. You can produce a Compliance Receipt at any time. Building out a Firm Masking Dictionary is a separate, opt-in workstream for your Firm Administrator. You do not need one in place to start working on real matters. There is no implementation project and no training period.

No. SafeIdea Platform sits beneath your product, not in front of it. It does not host content, does not deliver AI capabilities, does not own the attorney's relationship, and does not market into your installed base. The revenue model is per-seat compliance infrastructure for the attorney; it does not encroach on adjacent product revenue you generate.

The category SafeIdea owns, compliance infrastructure, is structurally separate from the categories you operate in: AI capability, practice management, legal-AI workflow, frontier inference. SafeIdea Platform is designed to expand your addressable surface, the matter types and document categories your customers can route through your product, without expanding your confidentiality liability.

Matter types and document categories your customers cannot route through your product today. The constraint is rarely your product's capability. It is the attorney's duty of confidentiality and the firm policy that flows from it. Matter-specific content gets routed manually, hand-masked, or kept off the AI surface entirely.

With SafeIdea Platform, confidential entities never leave the attorney's workstation; only sanitized prompts reach your service. The result is utilization growth from your existing installed base, in workflows you already support, where adoption was being held back by a confidentiality friction your product was never built to solve on its own.

Each SafeIdea Platform integration is configured against your product, not dropped in generically. The entity taxonomy reflects the entities that matter on your surface. The seed-source mappings, what we read from your product to seed Matter-Level Masking Dictionaries, are built around your data model. The canonical forms and the masking rules are tuned to your workflows and the matter types your customers run through you.

This is the same configurable dictionary-and-rules machinery that ships at the firm scope today; we apply it at the partner scope. Each connector is a distinct configuration. The Clio Connector, built and operational, is the template.

The boundary matters, so a clean answer. What is partner-shaped is the integration configuration: the taxonomies, seed-source mappings, canonical forms, and masking rules tuned to your product. That configuration does not transfer to a different vendor; if a firm later adopts a different platform, that platform requires its own SafeIdea integration with its own configuration.

What is attorney-owned travels with the attorney: Matter-Level Masking Dictionaries, Local Cross-Matter dictionaries, and Compliance Receipts live locally on the attorney's machine, and the Firm Masking Dictionary lives on firm storage the firm controls. The attorney is not held in place by SafeIdea. What holds value with you is the depth of partner-shaped integration that makes your product the most compliant surface for confidentiality-sensitive work.

Five preservations.

• Your customer relationship. Attorneys continue to engage with your brand, your workflows, your support. SafeIdea is not in front of you.
• Your contracts. SafeIdea Platform does not displace your DPAs, MSAs, or commercial terms with your customers. The compliance layer is a separate licensing relationship that does not require you to amend yours.
• Your AI vendor choices. SafeIdea is multi-LLM by design. Whichever model your product calls, the compliance layer is agnostic.
• Your margin. Customers maintain their own AI subscriptions. SafeIdea does not sit between you and any inference billing and is not an inference reseller.
• Your roadmap. Integration scope varies. The shape is what we discover together. SafeIdea does not push a capability roadmap onto your product.

No. SafeIdea is multi-LLM by design and agnostic to which model you or your customers use. If your product offers multiple AI options today, all of them continue to work. If your product calls one, that one keeps working. The compliance layer is downstream of which model is invoked and is not entangled with any model provider's commercial terms.

Three categories.

• Frontier AI vendors (Anthropic, OpenAI, Google as target categories). SafeIdea makes your models deployable in legal practice without your customers' confidentiality obligation surfacing as a deployment blocker.
• Legal-specific AI platforms (Harvey, CoCounsel, Vincent, Legora, Irys, Intapp). SafeIdea is the local compliance layer that lets your attorney customers use your platform on confidentiality-sensitive matters they otherwise would not route through it.
• Practice management and back-office (Clio, LawPay, ALPS, JD Supra, NetDocuments, Litera, state bar vendor programs). SafeIdea sits beneath your product and protects the customer's confidentiality obligation as their AI use scales across the workflows you already support.

The Clio Connector, in the practice management category, is built and operational today and is the proof point for what a partner-shaped connector looks like.

Two artifacts, both attorney-owned, both produced on the attorney's machine.

• Audited Session Records. Per-turn, immutable records of what was masked, what was sent, and what came back. Created locally as work happens.
• Compliance Receipts. Per-matter artifacts produced on demand by the attorney. Signed, cryptographically chained, tamper-evident. They never capture the underlying confidential content. They are what the attorney hands to a managing partner, ethics counsel, a malpractice carrier, a regulator, or a court if asked.

The compliance posture documented in a Compliance Receipt is attorney-level, not partner-level. Your partner liability surface is unchanged by the artifact's existence.

Yes, and this is the next step in the SafeIdea Platform partner motion. The SafeIdea Indexer already builds the Firm Masking Dictionary from named seed sources, with Clio one of them. Making a SafeIdea Platform integration a first-class seed source on the same Indexer pipeline turns your product from a route for masked content into a route for the masking knowledge itself.

The result for your customers is a shorter path to a firmwide compliance posture, and for your product a deepening role in the firm's AI compliance infrastructure. The Clio Connector seeds Matter-Level Masking Dictionaries today. Partner-fed Firm Masking Dictionaries are next.

Open to discussion based on integration scope and motion. SafeIdea (direct) carries a per-seat subscription at $399 per seat per month at list, with Charter Customer pricing at $249 per seat per month for attorneys admitted during the launch program. SafeIdea Platform accommodates several arrangements: revenue share on the SafeIdea seat, OEM, bundled distribution, or hybrid combinations. We start with discovery rather than a term sheet, because the right shape depends on which partner category you sit in and how the integration is going to be consumed.

The discovery call, thirty to forty-five minutes, founder to founder, is to understand fit: which partner category you sit in, what your customers' workflows look like, what your data model exposes, where your customers feel the confidentiality friction today, and what an integration shape would look like for both sides.

Integration timelines vary by scope. The Clio Connector is in production; subsequent connectors draw on what we learned building it. We give you a concrete timeline at the end of a discovery conversation, not before it.

The Partners page has a direct "Get in touch" form. Founder-to-founder discovery is the norm at this stage. Cart Reed (CEO) and Erich Wasserman (Founding Partner, CCO) will respond by email within a few business days to schedule the call.