Phaselaw is a document review and redaction platform from Phaselab, Inc. of New York, built for answering subject access requests, freedom of information and public records requests, and personnel records requests. It takes in documents, emails, chats, spreadsheets and images, removes duplicates and threads email, and lets a team mark documents in or out of scope with the reasoning recorded. Its own models suggest redactions across more than 50 categories of personal data, so that one redaction decision can be applied across a whole case.
An AI search feature turns a request written in plain language into a structured search the reviewer can see and edit. Exports include combined PDFs, individual files and a summary of the exemptions and review decisions applied. Customers shown include the NHS, HMV, YMCA and the Raspberry Pi Foundation, with pages for enterprises, privacy consultants and law firms, and the public sector. Data is hosted in the United States or the United Kingdom. Pricing is quoted by annual document volume and is not published.
Side by side on the same 15 capability axes and 12 legal signals: where the two part company, and what neither one publishes.
Other products that do the same kind of work as Phaselaw, and what each one does.
All 15 axes, graded from public sources on the date shown. Hover a grade to see what the letter means on that axis.
How much of the product is actually AI. Whether the machine learning is the mechanism the buyer is paying for or a feature layered onto conventional software, and whether the vendor is specific about which is which.
The models drive the core task inside a platform that would still work without them. Phaselaw's own models flag more than 50 categories of personal data as suggested redactions. An AI search feature added in August 2026 turns a request written in plain language into a structured search the reviewer can see and edit. Ingestion, deduplication, email threading, format conversion, redaction across a whole case, exemption tracking and export are document processing and workflow that a team could run by hand.
Whether the vendor publishes measured accuracy on citations and assertions, grounds output to primary sources, and says plainly what its system does when it does not know. Legal has a documented public record of fabricated citations reaching filed briefs, so an untested claim of accuracy is not evidence.
Accuracy is asserted, not measured. The customer pages say Phaselaw surfaces personal data automatically so nothing gets missed, and the changelog reports that name detection became significantly more accurate with fewer false positives. No detection rate, error rate, test set or comparison is published. What the models produce can be checked at the source: suggested redactions appear in place on the document for a reviewer to keep or redact, and the AI search shows the exact query it built. The product does not cite legal authority.
What the system decides on its own, what a lawyer must approve, and whether the vendor documents where the review point sits. A tool that drafts under review and a tool that files without one are different products and different risks.
The models suggest and a reviewer decides, with real review surfaces. A Suggested Redactions page lets a team keep or redact each item, or a whole category, across a case before opening documents; cases carry assigned reviewers and approvers, document statuses and resolvable comments; undoing a redaction across a case shows how many will change and asks for confirmation; and a Case Activity page logs who did what and when.
The vendor states that its AI features are clearly labeled and that their suggestions are displayed and editable by all users. Nothing states a limit on exporting before suggestions are reviewed, or what the product does when it misses personal data.
Named, dated evidence that the product works in production at real firms or legal departments. Case studies with figures and identified customers count. Unattributed testimonials and launch announcements do not.
Named customers, one with a figure, none measured. The home page shows logos including the NHS, Labor, Museum Wales, YMCA, HMV, the Raspberry Pi Foundation, the Guinness Partnership and Depaul, and quotes Tash Whitaker, DPO of Whitaker Solutions Ltd., on subject access work falling from 10 days to 3, and Paul Byrne, CEO of Propelfwd, a data protection consultancy in the Channel Islands, on subject access and freedom of information work.
The site's headline figures, 92% less time per request, over 95% of requests delivered on time and more than 10 million documents redacted, give no basis or date.
How client confidences are handled: attorney client privilege and work product treatment, segregation of one client matter from another, whether client data trains any model, and what the vendor commits to in writing rather than in marketing.
Strong commitments appear in marketing, and none in the published terms. The product and customer pages say each client gets isolated processing and data never crosses between cases. They say data stays in the region the customer chooses and is never shared with third party AI, and that access to cases is granted only where needed. The Terms of Use, which govern the product, contain no confidentiality clause, and no page says whether customer documents are used to train or improve Phaselaw's own models. Privilege and work product are not addressed.
Whether the vendor is clear that it supplies a tool rather than legal advice, who its audience is, and how it addresses unauthorized practice of law, competence and supervision duties, and jurisdiction limits. ABA Formal Opinion 512 is the reference point. Where the advice line is not the duty a product raises, the axis is read through the nearest professional duty it does raise: judicial conduct rules and the reviewing duty for products sold only to courts, and the duty to bill for time actually spent for products that draft time entries.
For a disclosure tool the professional question is review: whoever signs a disclosure answers for what it leaves in and what it takes out. The law firm page puts that question itself, asking what happens if something is missed when your name is on the disclosure, and answers that personal data is surfaced automatically so nothing gets missed. The Terms of Use provide the service as is and disclaim responsibility for the accuracy of results.
Nothing states that exemption and redaction decisions remain the reviewer's judgment, or that the product gives no legal advice.
Published governance over model behavior: who owns it inside the vendor, what is tested before release, and what is disclosed about disparate output across matter types, parties, or populations.
Principles are published, not a governance framework. The August 2026 release of the AI search feature says governance, security and privacy come first for all AI features, and commits to labeling AI features and showing their suggestions for users to edit. Administrators can switch individual detectors on and off. No accountable owner, testing before release, or finding on whether detection performs evenly across the kinds of personal data and documents it handles is published.
Retention, deletion, access control, and what happens to prompts and documents after they are processed. Whether the vendor states its subprocessors and its incident practice, or leaves the buyer to assume.
The product pages describe access to cases granted only where needed, single sign on over OAuth2, OIDC and SAML 2.0, SOC 2 Type II and annual third party penetration testing. The changelog adds a case activity log and optional protections against attempts to infer the text under a redaction. The Privacy Policy says personal data is kept as long as necessary to provide the service. Nothing states how long uploaded case documents are kept or how they are deleted.
No named subprocessor list or incident notification commitment is published outside the trust center, which is linked from the site footer.
What the vendor stands behind contractually when its output is wrong. Indemnities, caps, carve outs, insurance, and whether any of it is published or only reachable through a negotiated agreement.
Liability is addressed only through a standard limitation clause. The Terms of Use provide the service as is and exclude indirect damages, including those for accuracy of results. They cap Phaselab's liability at the greater of $100 or the fees paid in the twelve months before a claim, and require the customer to indemnify Phaselab against claims arising from its use of the service. Nothing stands behind a missed redaction or a redaction that goes too far.
How deeply the product reaches into the systems legal work already lives in: document management such as iManage and NetDocuments, Word and Outlook, contract lifecycle management, matter management, e-billing, and court filing systems.
Documents arrive by upload, including ZIP archives, PST and MBOX mailboxes and Slack exports, and the enterprise page says no integrations are required. No connection to a document management, case management, email or records system was found. Single sign on is supported for login, which is identity rather than a practice system.
Where the software runs and where the data sits. Multi tenant cloud, single tenant, private deployment, on premises, and whether region of residence is a published option or an enterprise conversation.
Region and processing are stated for a single hosted service. The pricing page includes US or UK data hosting in every plan, and customer pages say data is isolated in the region the customer chooses and never leaves that country. The August 2026 release says AI search queries are processed only in the country of deployment, inside Phaselab's own AWS environment. Each client's processing is described as isolated, without saying whether infrastructure is dedicated or shared, and no option on premises or in a private cloud is published.
Independent attestation a buyer can pull without a sales call: SOC 2, ISO 27001, penetration test summaries, a trust center with current reports and named scope rather than a badge image.
The product and customer pages state SOC 2 Type II certification, annual third party penetration testing and GDPR compliance. No auditor, scope or date is given on those pages. A trust center is linked from the site footer.
Which models sit underneath, whose they are, where they run, and whether the vendor commits to telling customers when that changes. A legal buyer inherits every dependency it cannot see.
The architecture is described without the models. The pages refer to Phaselaw's own models detecting more than 50 categories of personal data and say customer data is never shared with third party AI. The August 2026 release adds that AI search queries are processed only by Phaselab in its AWS environment, with no subprocessor added. No model, base model or architecture detail is named, and no commitment to tell customers when the models change is published.
Whether a buyer can learn what this costs without entering a sales process: published rates, the unit being charged, what sits behind an enterprise tier, and what implementation adds.
The unit and structure are published; the figure is not. The pricing page says price is based on the annual volume of documents processed, with no charge per user, unlimited seats, and full platform access, onboarding, training and support included with no add ons. The calculator on that page estimates savings rather than price, and every route to a figure is a quote or a demo. A pilot on the customer's own data is offered in place of a free trial.
Who the product is actually built for. AmLaw, midlaw, small firm and solo, in house departments, government and courts, and which practice areas are supported rather than merely claimed.
Buyers and request types are set out, short of the limits. The site addresses enterprise privacy, legal, HR and compliance teams, privacy consultants and law firms handling disclosures for clients, and public sector bodies, and covers subject access requests, freedom of information and public records requests, and personnel records, with US or UK hosting. It describes itself as built for subject access work rather than for litigation evidence review.
Supported file types are listed; volume limits, the languages and jurisdictions its detectors cover, and the requests it does not handle are not stated.
The public pages on file for Phaselaw, with the recorded signals each one supports and the date it was last read. Open any of them and check the reading against the record.
Prompt and Output Retention, Refusal and Uncertainty Behavior
Read Sep 22, 2026
Client Data in Training, Third Party Request and Subpoena Notice
Read Sep 22, 2026
Fabricated Citation Record
Read Nov 9, 2026
Billing and Fee Posture
Read Sep 22, 2026
Court Disclosure Support
Read Sep 22, 2026
Ethical Walls and Matter Segregation
Read Sep 22, 2026
Unit and structure published, figure withheld. Charged by annual volume of documents processed, per the pricing page, with tiers that scale with volume. No charge per seat: unlimited seats. Everything included: full platform access, all features, onboarding, training and support, with no fees or add ons. The calculator on the pricing page estimates savings from monthly requests, pages per request, time per page and reviewer wage, and shows no price.
It lists what is included: unlimited seats, AI redaction, support for emails, Teams, Slack, documents and images with OCR, single sign on, US or UK data hosting, round the clock technical support and unlimited training. The Terms of Use describe paid subscriptions billed through Stripe that renew automatically at the rate current at renewal. No rate, band, tier name or minimum is published.
Implementation: None published as a separate charge. The pricing page includes onboarding, training and support in the price, and the customer pages say most teams are running within a day with no integrations required.
Confidentiality and data terms: No Business Associate Agreement is offered or referred to on the site, and HIPAA is not mentioned. Health information can appear in the requests the product handles, and NHS bodies are among the customers shown; the site's data protection statements are framed around GDPR.
Note: No figure is published at any level. The pricing page offers a pilot on the customer's own documents instead of a free trial. The Terms of Use say subscriptions renew automatically unless the customer opts out by email, and that fees already paid for the current term are not refunded pro rata.
Phaselaw now surfaces black outs and improperly created redactions already present in uploaded documents as suggested redactions, so a reviewer makes an explicit decision on each one inside the redaction workflow. The same release lets case wide redactions match case, so redacting a term in capitals leaves the ordinary word alone.
Phaselaw added a Case Activity page that logs significant events in a case: case creation, status and deadline changes, uploads, and document status changes. It opens from a new icon beside the case title and is described as an audit log of who did what, and when. It only appears for cases created from 21 September 2026 onward, because it relies on newly stored case data.
Full page redaction was reworked so users select one or more pages to redact in a single action, mirroring area redactions, instead of redacting only the visible page. A new toggle on finalized exports excludes fully redacted pages so blacked out pages no longer appear in the export bundle. Both changes apply only to full page redactions made from 21 September 2026 onward, because of a change in how those redactions are stored.
A signal records what public sources say on the date shown. It is not a grade and it is not a recommendation. Where a signal reads Not addressed, it means the index did not locate the material in public sources on that date, which is a statement about disclosure rather than about the product.
Can material a lawyer puts into this product be used to train a model?
A published agreement or policy exists and none of it addresses the question either way, or the document that would answer it could not be read and the summary names the retrieval limit. The summary states which shape the silence takes: an improvement right granted that never names training, or no improvement right granted at all.
The Terms of Use and Privacy Policy are published, and neither says whether customer documents are used to train or improve Phaselaw's models. The Privacy Policy reserves a right to use aggregated, deidentified or anonymized data to analyze, build and improve the service. That right never names training and is framed around the personal data Phaselab collects as controller rather than the documents customers upload. The product pages say data is never shared with third party AI, which answers a different question.
How long does the product keep what a lawyer typed, and can that be set to zero?
A specific retention period is published and the customer cannot change it.
For the AI search feature released in August 2026, the vendor states that searches, results and intermediate processing steps are not logged or retained. That is the product's only free text AI input. No retention period is published for uploaded case documents, redactions or review decisions, which are part of the case record.
Does retrieval respect the firm’s ethical walls, or can the model read across them?
The product maintains its own permission model, documented, requiring the firm to keep it aligned.
Access is set per case in Phaselaw's own permission model, which the customer maintains; cases carry assigned reviewers and approvers. The law firm page says each case is separate, each client gets isolated processing, and data never crosses between cases, so a firm can work for several clients at once. No inheritance of permissions from a document management system is described.
If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?
Published terms or policy address disclosure to authorities or in response to legal process, and no commitment or reservation regarding customer notice is located anywhere. The vendor has told the customer that data can leave and has said nothing about whether the customer hears of it.
The Privacy Policy lists meeting legal obligations, including court orders and other legal process, among the purposes for which Phaselab may share personal data with third parties. Nothing addresses whether the customer is told of such a request, and nothing addresses requests for the case documents a customer uploads, for which Phaselab describes itself as processor.
Where does the law in this product come from, and does the vendor have the right to use it?
No located public material identifies the corpus behind the product’s answers.
Provenance questions narrow to the customer's own uploads and exemption choices. The product works on the documents each customer uploads. Exemption reasons are chosen by the customer, and no body of law behind them is identified.
Does the product tell you when the authority it just cited has been overruled?
No located public material addresses whether authority is checked for subsequent history.
Redaction and disclosure work gives treatment checks nothing to attach to. The product does not cite legal authority, and nothing addresses checking authority for subsequent history.
What does the product do when the answer is not in the corpus?
The vendor describes refusal or abstention behavior in public materials.
The August 2026 release of the AI search feature describes two behaviors. When asked for a search the system does not support, it returns an error explaining the problem instead of a guess. When it translates an ambiguous request, it shows a warning explaining what it did. Both are illustrated with screenshots. No confidence indicator or equivalent behavior is described for the models that suggest redactions.
Does a public court record exist addressing fabricated or hallucinated legal citations in output from this product?
No court order, opinion or disciplinary record addressing fabricated or hallucinated legal citations produced by this product has been located as of the date shown. This is a statement about the public record on that one subject, not a finding about the product, and this signal is not a litigation history.
The AI Hallucination Cases database kept by Damien Charlotin holds no entry naming Phaselaw as the tool used.
Has the vendor engaged in public with the ethics opinions its buyers are bound by?
No located public material engages with bar or ethics guidance.
Data protection and public records law frame the product, while lawyers' own professional duties sit outside that frame entirely. The site refers to data protection law, the Information Commissioner's Office and public records laws, but no material engages with lawyers' professional or ethical obligations or names any ethics opinion.
Does the vendor address what happens to the bill when the work takes an hour instead of six?
Public materials claim time savings without addressing billing or disclosure, and the product sits inside a fee relationship between a lawyer and a client where those savings would change the bill.
The law firm page claims 92% less time per request and frames the saving as protected margin on work for a fixed fee. Nothing addresses work billed by the hour or what a client is told about review assisted by AI. For the enterprise and public sector teams that make up much of the customer base, no client is billed for the work.
Can a firm get this vendor through a client’s AI clause without a bespoke negotiation?
No located public material supports a client side disclosure obligation.
No subprocessor or model provider list is published outside the trust center: the Privacy Policy names categories of service provider and Stripe, and the changelog names AWS. The trust center is linked from the site footer.
If a judge’s standing order requires an AI disclosure, can the product produce one?
Some elements of the record are available, short of a document level export.
No court sits in the path of most of this work; the body that asks is the regulator. The enterprise page says that if the Information Commissioner's Office challenges a response, every decision is logged, showing what was found, who reviewed it and why each redaction was made. Exemption reasons are tracked per redaction, and a Case Activity page logs activity from September 2026. A spreadsheet summary export records exemptions applied, duplicates, redundant emails, decisions that documents are out of scope, and errors.
Files marked out of scope show whether a person or deduplication removed them. Nothing records which redactions began as model suggestions.
Every grade and every signal on this index is drawn from public sources and dated. If a record is wrong, out of date, or missing an artifact the index did not locate, send the source and it will be reviewed and the record redated. Vendors are welcome to submit documentation. Nothing on this index is for sale, including a listing, a placement, or a grade.