AI Disclosure
Effective September 29, 2026
This AI Disclosure explains how Speider uses artificial intelligence in its product, what the AI does and doesn't do, the limitations subscribers and their clients should know about, and the responsibilities that fall on each side of the platform. It supplements the Privacy Policy (which covers data handling) and the Terms of Service (which is the contract between you and Speider); it doesn't replace either.
Read this if you're a subscriber, if you advise a client whose work product touches Speider output, or if you're evaluating Speider for adoption.
Speider is a legal intelligence monitoring platform for lawyers advising business clients. The product takes a structured profile of each client (counterparties, jurisdictions, agreements, regulatory exposure, monitoring keywords), compresses that profile into a "Client Lens," and uses the Lens to find and rank publicly available material from external sources. The output is a briefing, which you start with the Generate Pull button: ranked articles, summaries, contextual notes about why each article matters to that client, and considerations for the lawyer.
The specific places in the product where AI is used are described in Section 4.
Speider uses commercial AI services from two providers:
Neither provider uses Speider inputs or outputs to train its models, under our agreements with each. Model identifiers are pinned in our codebase and updated deliberately when we choose to update them.
Speider uses AI in nine places. For each, we describe what goes in, what comes out, what the output is for, and what it isn't.
When you enter a company name on the new-client intake form, Speider runs web searches and pre-fills the rest of the form: public or private status, ticker, industry, jurisdictions, suggested counterparties, regulatory agencies, named individuals, and likely agreement types. Sources are cited.
You can edit any field before saving. Nothing the autofill produces is final until you save the client profile.
If none of the built-in client templates fits, you can describe the kind of client in a few words ("boutique hotel operator") and Speider drafts a generic starting point for the intake form: industry, monitoring keywords, agreement types, regulators, and a monitoring focus.
The draft names no real company or person, and it fills the form rather than saving anything. You review and edit it before the client profile is saved.
When you save a new client profile or regenerate an existing lens, Speider compresses the structured intake data and any uploaded documents into a Client Lens: a privacy-safe list of names, flags, monitoring keywords, and obligation criteria used to find news for that client.
The Lens is editable from the Client Lens page. You can remove individual values, add focus prompts, and create additional lenses scoped to different practice areas for the same client.
When you click Generate Pull, Speider gathers recent material from its sources and rates each item's relevance against the Client Lens. The ranking and the relevance dots on each card come from those ratings, and the AI may add a sentence of context to the briefing's summary line.
Every article links back to its original source. Click through and verify before relying on any factual claim.
Once a briefing is saved, Speider writes a summary, context for your client, and considerations for the top-ranked cards, the ones you're most likely to read. Each is written from that article's stored text and the Client Lens.
A sentence the article's text doesn't support is removed rather than shown. That lowers the risk of an invented fact; it doesn't remove it, so read the source before relying on a card.
When you export a briefing as a Word memo or an email draft, Speider writes the memo's prose for the articles you selected, from each article's stored text and the Client Lens, under the same rule as card summaries.
The export is a draft for you to edit. It goes nowhere until you send it yourself.
From a single article card, Speider can draft a short note to your client about that development. The note reports facts from the article and raises questions for the client; it doesn't draw legal conclusions or recommend a course of action.
If any fact in the draft can't be checked against the article, no note is produced. You edit the note and send it under your own name.
Scout is the in-product chat assistant. He answers questions about how the product works, helps you understand a briefing, and stays on the product itself.
Scout doesn't give legal advice. His instructions tell him not to take political or social positions and not to weigh in on whether a development is good or bad for a client. If you ask Scout a legal question, treat his answer as a starting point and check the law yourself.
When you upload a contract, relationship matrix, or list (PDF, Word, or Excel) on the intake form or the edit-client form, Speider transmits the file to Anthropic for a single parse call. Anthropic returns a narrow set of structured fields: the client's industry, counterparty names, agreement types, jurisdictions, the regulatory agencies the document mentions, obligation types, and suggested monitoring keywords.
The document is held in serverless function memory for the duration of the parse call (typically under thirty seconds) and discarded the moment the call returns. The original is never written to our database or to disk on our infrastructure. From uploaded documents, we never extract the names of individuals (not executives, not signatories, not anyone), dollar amounts or payment terms, dates or deadlines, contract language, trade secrets or technical specifications, or personnel and contact details. See Section 7 for the data flow.
Speider's AI isn't a lawyer, and it isn't your lawyer. In particular:
Subscribers and their clients should understand the following limitations before relying on Speider output:
What you put in (the client profile, uploaded documents, additional monitoring context) is stored in our own database (Supabase) and processed by our two AI providers, Anthropic and Voyage AI. Search terms built from it also go to the news and court search services named below. The full data-handling description, with every vendor, is in the Privacy Policy. The AI-specific points are:
For full data handling, including retention, vendor list, breach notification, and your rights as a data subject, see the Privacy Policy.
By using Speider, you agree to apply the following responsibilities to every briefing, summary, consideration, and Scout response you read:
Whether you must affirmatively disclose AI use to a client depends on your jurisdiction, the nature of the matter, and the role the AI-generated material plays in your work. Several state bars have published guidance addressing this question:
This list isn't exhaustive. Check your own jurisdiction's most current guidance.
The following template is a starting point for a client-facing disclosure. It isn't legal advice, it isn't jurisdiction-specific, and it should be reviewed against your bar's current rules before sending. You're responsible for the final form and content of any disclosure you make to your client.
Dear [Client],
Part of the monitoring work I do for you relies on Speider, a legal monitoring tool that uses AI. The tool checks publicly available material against a profile I maintain for you: SEC filings and press releases, federal and state court decisions, the Federal Register, FTC and DOJ press releases, product recalls, workplace-safety notices, patent filings and challenges, class-action filings, the legal trade press, and business news. It then produces a briefing that I review.
The profile is kept in the tool. It holds what I enter about your business, such as your industry, the companies you deal with, the kinds of agreements you have, and the regulatory areas I watch. If I upload a document to build that profile, the tool reads it once to fill in those details and doesn't keep the document itself. The tool's AI providers don't use any of this to train their models.
The AI doesn't provide legal advice and doesn't make decisions about your matter. I review every briefing it generates, verify the underlying sources before relying on anything material, and apply my own professional judgment.
If you have questions about how I use AI in your matter, or if you'd prefer I not use AI tools for any part of your work, please let me know.
Sincerely, [Your name]
We may update this AI Disclosure when:
For material changes, we'll notify subscribers by email or by an in-product banner before the change takes effect. The effective date at the top of this document reflects when the current version was last revised. Continued use of Speider after the effective date of a revision constitutes acknowledgment of the revised disclosure.
Questions about this AI Disclosure, the AI models we use, or your obligations in using Speider:
Questions? Email hello@speider.ai or read our Privacy Policy.
Try it free