Fileread
Fileread is a legal AI platform for litigation, eDiscovery and investigations. It reads the entire production, including formats other tools cannot open, and returns every answer cited to the exact source page, so verification takes seconds.
What is Fileread?
Fileread is an end-to-end legal data intelligence platform built by Fileread Inc., a New York company founded by Chan Koh (CEO), Daniel Hu (CTO) and Freya Zhou (COO). It positions itself across the whole life of a matter, from the first production to trial, and its promise is deliberately narrow: read the entire record, including the formats other tools cannot open, and cite every answer back to the page it came from, so checking a fact takes a glance instead of an afternoon.
The product is organised around four named modules. Insights answers natural-language questions across the full matter, with a citation on every assertion. Natural language boolean search turns a plain description into a precise, reviewable query rather than a fragile boolean string. Axis pulls structured tables out of unstructured productions, each cell cited. Workbench, announced in July 2026, is the agentic layer: it plans and runs long tasks, builds its own searches, executes repeatable procedures and assembles the work product. Matter Intelligence, released in November 2025, reads the foundational documents of a case and generates themes and queries in under five minutes, removing the prompt-engineering barrier; a Deep Search Agent shipped in June 2025.
The stated doctrine is breadth plus precision: cover the whole volume and the mix of formats - text, images, handwriting, unstructured data - while pinning every assertion to its exact location. The published examples are concrete: a production that arrived as 4,000 SMS screenshots was converted and de-duplicated into 725 real messages, saving more than twenty attorney hours; a trade-secret investigation ran across more than 100,000 emails.
Fileread does not ask teams to move their data. It runs on the productions already in hand, and its native, real-time Relativity integration, with a Code in Relativity button since December 2025, lets firms keep their existing repository. It has also been sold through Microsoft Azure Marketplace since October 2025, with unified Azure billing. Security is presented as a precondition rather than a footnote: SOC 2 Type II, ISO 27001, HIPAA and GDPR, encryption in Azure, private LLMs. Named references include Proskauer Rose LLP, Smith Gambrell & Russell and Troutman eMerge.
What it does
- Ask a question about the whole production in plain English and get an answer cited to the page it came from
- Verify any claim in one click, straight through to the source document page
- Assemble the work product: fact memos, chronologies, contradiction reports, cast-of-characters lists and org charts
- Extract structured tables from unstructured productions, with every cell cited
- Open formats other tools cannot read: images, handwriting and message screenshots
- Identify custodians, key dates and gaps between the allegations and the production from day one
- Code documents in Relativity without leaving the workflow
When to use Fileread / When not to
A quick filter to help you decide if Fileread is the right fit.
When to use Fileread
- Litigation teams working through large document dumps who need the decisive facts pinned to a page, from early case assessment through deposition prep to trial support
- eDiscovery and litigation support managers handling heterogeneous productions - text, images, handwriting, SMS screenshots - that other review tools cannot open
- Firms already standardised on Relativity that want an analysis layer on top of their existing repository, with no data migration
- In-house legal and compliance teams bringing investigations back in house, such as a trade-secret review across more than 100,000 emails
- Alternative legal service providers and litigation support vendors, including named partners such as JND Legal Administration and Cimplifi
When not to use Fileread
- Solo practitioners and cost-sensitive small firms: no price is published, there is no free plan or trial, and the only entry point is a sales conversation
- Individuals: the vendor states plainly that it provides its services to organisations, not individuals
- Anyone needing a mobile app, a browser extension or a documented public API - the product is web-only and no API documentation is published
- Teams whose matters are not in English or not shaped by US discovery practice: no interface or processing language beyond English is declared, and the workflow assumes Relativity, ESI and document productions
- Organisations looking for general-purpose document management or enterprise knowledge search: Fileread is built for legal productions and case records, not for everyday business files
How to use Fileread
A typical end-to-end flow, from setup to results.
- Start with the contact form on /letschat - See it on your matter, Get a demo or Talk to sales; there is no self-service sign-up
- Watch the interactive walkthroughs on the product page beforehand if you want context before the call
- Choose the deployment: work end to end inside Fileread on a live production, or connect Relativity and keep your existing repository
- If you buy through Microsoft Azure Marketplace, purchase on your existing Azure account and take the charge on your Azure invoice
- Point Fileread at the productions you already hold - there is nothing to migrate
- Upload the foundational documents (complaint, memos, key exhibits) so Matter Intelligence returns suggested themes and queries in under five minutes
- Edit the generated queries and summaries, keep several contexts for the different facets of the matter, and roll back through version history
- Ask questions in plain English, or describe the search you want and let the natural language boolean converter build it
- Verify every claim by opening it at the exact page of the source document
- Assemble the work product in Workbench, and code documents in Relativity from the same workflow when the integration is connected
Pros & Cons
Pros
- Every answer is cited to its source page, so verification takes a glance rather than an afternoon
- Reads formats other tools cannot open - images, handwriting, message screenshots - and normalises them: 4,000 SMS screenshots became 725 de-duplicated messages, saving more than twenty attorney hours
- Adds to the existing workflow with no migration: it runs on productions already in hand, and the native real-time Relativity integration lets teams keep their review platform
- Serious, documented compliance base: SOC 2 Type II, ISO 27001, HIPAA, GDPR and CCPA, with audit and penetration test reports available through the trust center
- Explicit commitment that customer data never trains models, backed by a published subprocessor list and an available DPA
- Named, checkable references at well-known firms - Proskauer Rose LLP, Smith Gambrell & Russell, Troutman eMerge - and coverage in the New York Times, TechCrunch and Law.com Legaltech News
- Sustained delivery pace, with around a dozen features announced between May 2025 and July 2026, and simplified procurement through Azure Marketplace for organisations already on Azure
Cons
- No public pricing at all: no pricing page, no amount, no tier - budgeting requires a sales conversation
- No free trial and no free plan announced; every call to action leads to the demo form
- The published privacy policy covers the website only - the commitments on case data live in a separate, non-public agreement
- GDPR compliance is claimed, but no Article 27 EU representative is named, no legal basis is stated and no standard contractual clauses are mentioned
- No published data retention period, and no hosting region named beyond the mention of Azure; the vendor declares it operates in the United States
- No mobile app, no browser extension and no documented public API
- Firmly aimed at US litigation - Relativity, ESI, productions - with little sign of adaptation to other legal systems, and no interface language other than English declared
Pricing & Plans
Fileread publishes no pricing. There is no pricing page on the site, none in its sitemap, and targeted searches, including legaltech directories, returned no figure. No free plan and no free trial are announced, and every call to action leads to the /letschat contact form, so pricing is quote-based. A second procurement route exists for organisations already on Microsoft Azure: Fileread has been available on Azure Marketplace since October 2025, purchasable on an existing Azure account with charges consolidated on the Azure invoice. Any amount quoted elsewhere should be treated as unverified.
- Fileread discloses no per-seat
- per-user or per-volume rate card
- scope and price are agreed with the sales team through the contact form
- purchase on an existing Azure account
- with charges consolidated on the Azure invoice
Data, GDPR & hosting
A consolidated view of how Fileread handles your data.
GDPR overview
GDPR compliance is claimed as a badge rather than documented in the privacy policy. The security page states the platform meets the highest compliance standards, including SOC2 Type II, ISO 27001, HIPAA compliance and GDPR requirements, and the trust center lists GDPR and CCPA alongside SOC 2 Type 2 and ISO/IEC 27001:2022, with a Data Processing Agreement under its Legal section. The policy actually published, last modified 1 May 2025, is a US website policy: it never cites the GDPR, states no legal basis, mentions no standard contractual clauses and names no Article 27 EU representative. It confirms the site is operated in the United States and that personal data may be transferred there. No DPO is named; requests, limited to data collected through the website, go to privacy@fileread.com. Detailed compliance documentation is released on request via trust.fileread.com.
Who owns the data?
Ownership of case data is not settled on the public site. The published privacy policy is explicit that it covers the website only, and states that Fileread's obligations to a client organisation are governed by a separate written agreement - the Master Services Agreement available through its trust center - so contractual review is unavoidable. On the website side, Fileread owns its own content and grants visitors personal, non-commercial use only; anything sent unsolicited (Submissions) carries a worldwide, perpetual, irrevocable, sublicensable, non-exclusive licence to Fileread, without compensation. Fileread states it does not and will not sell personal information, but reserves the right to transfer data in a merger, acquisition, bankruptcy or change of control.
Reuse rights
Website content (the Fileread Content) may be used for personal, non-commercial purposes only; reuse, reproduction or redistribution requires permission. What a customer may do with the outputs generated on its own matters is not addressed publicly - that sits in the separate service agreement. On the vendor's side the commitments are clearer: nothing uploaded trains a shared model, the system does not train on case data, and client information is never used to improve models or shared across matters. Data is encrypted in Azure and processed through private LLMs. The trust center discloses the subprocessors involved: Langfuse (prompt logging), OpenAI, Baseten, Pinecone and Microsoft Azure. Website analytics are handled separately: log data (IP address, browser, device, page requested), cookies, web beacons, Google Analytics and retargeting cookies are used, analytics are shared with partners only in de-identified and aggregated form, and the site does not respond to browser Do Not Track signals although it honours standard opt-out mechanisms.
Data retention & training
Hosting summary
Fileread states repeatedly that data is encrypted in Azure and processed through private LLMs - on the home page, on the corporate legal solutions page and on the security page, which describes military-grade encryption within Azure. Microsoft Azure is among the subprocessors disclosed at the trust center, alongside Langfuse, OpenAI, Baseten and Pinecone; AWS also appears on the infrastructure side. Beyond the platform name, nothing is specified: no Azure region is named, no country of residence is given for service data, and no data residency option is documented. The only jurisdictional statement in the published documents concerns the website: the privacy policy confirms that Fileread maintains and operates its website in the United States and that personal data may be transferred to and held on computers outside the user's own jurisdiction, where privacy laws may be less protective. Read together, the United States is the reasonable working assumption, but it is inferred from a website policy rather than declared for the service. Any buyer with a data residency requirement should have it confirmed in writing in the service agreement.
Things to keep in mind
Risks and trade-offs to weigh before adopting Fileread.
- No public pricing: budgeting, benchmarking or comparing means entering a sales process, with no trial to test the fit beforehand
- The published privacy policy covers the website, not the service. The commitments on case data are contractual and non-public, so the security badges on the marketing pages cannot replace a reading of the Master Services Agreement
- A citation makes verification easy, which makes it tempting to skip. A pin to a page proves where a sentence came from, not that the inference drawn from it is right; a team that stops opening the source will eventually carry an error into a filing
- Delegating first-pass fact-finding to an agent can quietly erode the close reading through which junior lawyers learn a case. Keep people reading, not only checking
- GDPR compliance is claimed by badge, but no Article 27 EU representative is named, no legal basis is stated and no standard contractual clauses are mentioned - a real gap for a European buyer
- No retention period is published, the vendor operates in the United States and may transfer personal data there, and OpenAI appears among the declared subprocessors: all worth checking against client confidentiality undertakings and any protective order
- Audit evidence (SOC 2, ISO, penetration tests) is released only on request through a trust center that blocks automated access, and the two legal documents carry different revision dates depending on the page hosting them - 1 May 2025 on the terms page, 22 July 2026 in the copy embedded in the privacy page
Setup & Integrations
Technical difficulty
Low, once the commercial step is done. There is no self-service sign-up: everything starts with a demo request. After that, no migration is required - Fileread works on the productions already held, and the Relativity connection imposes no change to existing processes or coding taxonomies. Organisations already on Azure can deploy faster through the Marketplace on their existing infrastructure. Matter Intelligence removes the prompt-engineering barrier by generating themes and queries from the foundational documents in under five minutes, and the vendor claims a zero learning curve with immediate impact.
Deployment
Integrations
Behind Fileread
Fundraising
Social
Resources
All the official URLs gathered for verification and reference.
Frequently asked questions
What is Fileread?
Does Fileread train its models on our case data?
Which formats can it handle?
Do we have to migrate our data?
Does it integrate with Relativity?
Which certifications does Fileread hold?
Is a DPA available, and who are the subprocessors?
How much does Fileread cost?
Is there a mobile app or a public API?
Who is behind Fileread?
Should you pick Fileread?
Fileread occupies a narrow position and occupies it well: document-heavy litigation, where the problem is not summarising but finding and proving. Its central argument, every answer cited to the page it came from, is also its real differentiator, and it is why the tool fits a deposition, an impeachment analysis or a trial binder rather than a general research workflow. The compliance base is above the sector average: SOC 2 Type II, ISO/IEC 27001:2022, HIPAA, GDPR and CCPA, a Data Processing Agreement, a published subprocessor list, and audit and penetration test reports released on request. The commitment that customer data never trains a model is stated explicitly and repeatedly.
Two reservations deserve weighing before a purchase. The first is complete pricing opacity: nothing is published, no plan is named, and there is no trial to fall back on, so every evaluation starts with a sales conversation. The second matters more to a legal buyer: the privacy policy actually published, last modified 1 May 2025, covers the website only. The commitments that count - how case data is handled, retained and deleted - live in a separate Master Services Agreement that is not public. That is a normal enterprise arrangement, but it means the security badges on the marketing pages cannot substitute for contractual review.
The company itself is small: a 6 million USD seed round led by Gradient Ventures in 2023, no later round found, three co-founders, and named clients in serious firms. Against that, the delivery pace across 2025 and 2026 is sustained, and the Relativity and Azure Marketplace routes lower the practical cost of adoption. For a US litigation or eDiscovery team already sitting on unmanageable productions, Fileread earns a demo.
- Choosing a selection results in a full page refresh.
- Opens in a new window.