D
Deep Legal
Deep Legal, a Starian company based in São Paulo, sells Quantum, a litigation analytics platform for Brazilian legal departments and firms with ten modules under one login. Its models have read and calculated more than one million cases, and Renner, PicPay, Braskem, Mercado Livre and Banco Safra are among its customers.
Jurimetria reads trial and appeal decisions claim by claim and turns them into 160 indicators, from outcomes and settlements to time to judgment, by court, city, judge, subject and law firm. More than 90 signals each carry the source excerpt. Prediction estimates each case's chance of loss, moral damages and decision date, rates exposure as probable, possible or remote, and shows forecasts against actual results. Benchmark compares up to three portfolios, including a competitor's built from its company registration number, and flags outliers and predatory litigation. A report agent answers questions with evidence from up to 3,000 decisions.
Legal Ops reconciles the buyer's portfolio with court records and writes closures back into its legal ERP. Grupo SADA uses Deep Legal for decision analysis beside its 2,482 cases in Projuris Empresas. Subscriptions are quoted, and single reports can be bought online.
Alternatives to Deep Legal
Other products that do the same kind of work as Deep Legal, and what each one does.
- LLexbeLexbe is a cloud eDiscovery platform for law firms, corporate legal departments and managed review providers, sold by Lexbe Inc. of Austin, Texas, which was founded in 2006.
- EEpiq AIEpiq AI is the AI product line of Epiq Systems, Inc., a New York legal services and technology company whose other businesses include class action and bankruptcy administration, managed…
- AAid4Mail InvestigatorAid4Mail Investigator is the email forensics and eDiscovery edition of Aid4Mail, Windows software from Fookes Software Ltd of Charmey, Switzerland, which has made email tools for more than 25 years.
- BBriefCatchBriefCatch is legal writing and citation checking software for lawyers, judges and courts, used in Microsoft Word, Outlook or a browser.
- OOpenText eDiscoveryOpenText eDiscovery is an electronic discovery platform covering collection, processing, analysis, review and production, sold to corporate legal departments, law firms and investigating…
- TTheta LakeTheta Lake captures, archives and supervises the communications regulated firms run on, across voice, video, chat, mobile messaging, email and whiteboards, through more than 100…
Capability grades
All 15 axes, graded from public sources on the date shown. Hover a grade to see what the letter means on that axis.
AI Centrality
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.
Portuguese language models read each claim in trial and appeal decisions and extract the outcome, reasons, amounts and deadlines. Predictive models estimate loss, moral damages, decision dates and settlement odds from a company's litigation history, and 12 month event forecasts arrive as ranges. A dashboard explains the portfolio in natural language, a report agent answers questions about its results, and reports can be turned into presentations.
The Intelligence module adds AI case law research and lookups of prospects and suppliers. Beside the models sits a conventional monitoring and reconciliation workflow that compares open matters with court status, shows discrepancies and updates the legal ERP.
Citation Accuracy and Hallucination Disclosure
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.
Jurimetria reads trial and appeal decisions claim by claim, and the case view carries the extracted result, reason and basis for each claim. More than 90 signals each show the original excerpt from the court publication, and the report agent answers questions over up to 3,000 decisions with evidence from individual matters. The settlement strategy report draws on similar decisions already handed down in the same court.
Deep Legal targets data extraction on more than 90% of decisions and sets a 90% global service level for analytics, and the predictive screen shows estimates against actual results. It reports no measured error rate for the agent's answers or summaries.
Autonomy and Oversight Model
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.
Portfolio checks against court data flag missing or closed cases and update the customer's legal ERP when records change, raising alerts when someone needs to act. Liaison staff see each discrepancy, the court status behind it and the resulting update. Forecasts sit beside the case information as estimates, with ranges and a comparison to actual outcomes. The contract makes the customer responsible for monitoring and checking information, and Deep Legal positions the software as support for the lawyer who handles the case. No forecast confidence level sends a case to counsel automatically.
Operational and Outcome Evidence
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.
Grupo SADA uses Deep Legal to analyze court decisions, follow trends by subject and benchmark its outcomes against the market. Its case study keeps that work apart from case storage and provisioning in Projuris Empresas, where the group structured 2,482 active cases. Quantum's customers include Renner, PicPay, Cielo, Stone, Serasa, JBS, Sabesp, Braskem, Neoenergia, Mercado Livre, Banco Safra, Mapfre and Telefônica. Each deployment tracks closed cases found, discrepancies and provisions beside forecasts, so a buyer can compare results over time. The Grupo SADA case gives scope and use rather than a measured change in forecast accuracy.
Privilege and Confidentiality Posture
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.
Each customer chooses which portfolio enters the platform, and access runs on individual credentials and roles. Confidentiality lasts for 24 months after the contract ends. The terms let Deep Legal and its partners use information the customer supplies to improve the software and prepare general statistics without identifying the customer, and the personal data addendum prefers anonymization and limits service providers.
The analytics combine official court data with the private portfolio information the customer adds. Attorney work product receives no treatment of its own in the model pipeline.
UPL and Professional Responsibility Posture
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.
Case tracking and analytics support the law firm or department responsible for the underlying dispute. The terms make the customer verify generated information and leave business and legal decisions with it. Court source limits are spelled out for case monitoring, including some sealed and physical cases, and Deep Legal says the software does not replace the lawyer's own tracking and work. That boundary covers alerts and forecasts alike.
Counsel acting on a forecast or on an answer from the report agent receive no supervision guidance tied to a jurisdiction.
AI Governance and Bias Disclosure
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.
Portfolio models show forecasts beside actual outcomes case by case, and Deep Legal sets a target of raising prediction accuracy every quarter. It warns that estimates widen for small or mixed portfolios and aims prediction at civil portfolios with more than 1,000 cases of history. Its responsible AI guidance classes risk by workflow and applies automatic anonymization and privacy by design to sensitive data. The Civil and Labor dashboards let a legal team read outcomes by court or judge, and Benchmark flags outliers and predatory litigation.
Deep Legal names no person accountable for model assurance and has released no test of how forecasts differ across claimant groups.
AI Safety and Data Stewardship
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.
Daily backups are kept for seven days, so a backup is unusable from day eight. Access data is deleted within 45 days of termination, and after 60 days Deep Legal may remove the remaining data, subject to its legal obligations. Users sign in with individual credentials, with Active Directory as an option, and roles decide what each sees. The data processing addendum sets 72 hour notice for certain personal data incidents, limits subprocessors and gives a removal process for personal data.
Sensitive data is anonymized automatically under the company's privacy by design approach. Prompts to the report agent and its answers follow the same platform schedule, with no separate handling period.
AI Liability and Recourse
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.
The customer terms cover proven direct harm caused by a material breach of an essential obligation. General liability is capped at the fees paid in the prior 12 months, or the amount paid under a shorter contract. Deep Legal defends and indemnifies the customer against certain third party claims that its software infringes intellectual property, provided the customer gives notice and lets it run the defense. The service is licensed as is, and the customer verifies the results. A wrong forecast, a misread decision or a business call made on the analytics carries no recovery beyond those terms.
Practice Systems Integration Depth
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.
Legal Ops reads portfolio cases from the courts and writes closures or updates back into the customer's legal ERP. The customer loads a spreadsheet or brings data through the integration. The API sends and receives case data through web services and JSON, and the organization chooses the fields it needs and configures authentication with its systems. The standard API is part of the subscription, while custom formatting or a new connector can need separate work.
Onboarding follows a 90 day plan. Cases are uploaded in the first week, finished cases are closed and the ERP cleaned in weeks two to four, analysis runs in month two, and ERP writes, alerts and prediction switch on in month three. Lawyers can also export a selected set of dashboard data to Excel.
Deployment Model and Data Residency
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.
Customers use Quantum online at an address agreed in the subscription. Deep Legal runs the hosting on AWS and DigitalOcean, and accounts use individual credentials with an Active Directory option. The data addendum allows cross border processing for purposes listed in the customer contract. The platform is built around Brazilian law and the court documents it reads. Deep Legal lists no hosting regions and does not say where generative processing runs relative to where case documents are stored.
Security Certifications and Trust Center
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 hosted system takes daily backups, uses individual credentials and connects to Active Directory where needed. Deep Legal keeps a data protection portal with its terms, privacy policy and processing agreement, and the addendum covers incident notices, personal data safeguards and a limited remote audit route for customers. AWS and DigitalOcean supply the cloud infrastructure. Deep Legal holds no audit report or certification whose scope covers the Quantum service, its hosting and its AI processing.
Model Supply Chain Disclosure
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.
Natural language processing components adapted in house read Portuguese court documents, on a largely open source stack that processes case data, applies the predictive models and serves dashboards. AWS and DigitalOcean host it. The customer terms separate third party services from the software license, and the addendum allows outside AI services for defined purposes, such as delivering and improving the service. The model behind the generative dashboard, the report agent and the AI case law research goes unnamed, along with its provider, and no notice is promised when it changes.
Commercial Transparency
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.
Deep Legal sells a continuing Quantum subscription, with a minimum one year term, and one off purchases of a defined report or analysis. Its online data store sells jurimetric and civil and labor reports by company registration number, paid online, and custom questions are priced on request. The standard API is included, while new integrations or formatting can cost extra, and implementation consulting and training are contracted separately.
Onboarding runs about 90 days, and Jurimetria setup takes up to 10 business days once the buyer's data arrives. Subscription prices come from a proposal.
Firm and Practice Coverage
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.
Quantum serves organizations with Brazilian civil and labor portfolios: legal and finance teams tracking large caseloads, and firms that want comparative court data. Case capture covers first instance, appellate and superior courts. Standard Jurimetria collection leaves out physical cases, criminal and administrative proceedings, sealed cases, matters outside the national CNJ numbering standard and cases before 2008.
A separate module tracks tax disputes before CARF, the federal administrative tax appeals council, and the Intelligence module runs nationwide lookups by name or company registration number. Forecasts widen for small or mixed portfolios, and the civil tool works best above 1,000 cases.
7 public documents
The public pages on file for Deep Legal, 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.
-
Client Data in Training, Ethical Walls and Matter Segregation, Third Party Request and Subpoena Notice
Read Nov 20, 2026
-
Bar Guidance Alignment
Read Nov 20, 2026
-
deeplegal.com.br/jurimetria1 signal
Primary Law Corpus Provenance
Read Nov 20, 2026
-
deeplegal.com.br/legal-ops1 signal
Billing and Fee Posture
Read Nov 20, 2026
-
deeplegal.com.br/loja1 signal
Court Disclosure Support
Read Nov 20, 2026
-
Outside Counsel Guideline Readiness
Read Nov 20, 2026
-
Fabricated Citation Record
Read Nov 20, 2026
No published figure
- Deep Legal sells a Quantum subscription with a one year minimum, or single reports and analyses bought online or by proposal. The standard API comes with the subscription. Custom integration, training and implementation can cost extra, and onboarding runs about 90 days.
Quotes cover the subscriptions and modules selected. The service guide also describes a scoped purchase for individual reports or temporary work. The customer terms say the scoped license ends when the deliverable is complete.
Implementation: The terms allow separately contracted setup consulting and training. Jurimetria setup can take up to 10 business days after the required data is delivered. Custom API connections and automation can require a separate proposal.
Note: The recurring license renews automatically unless the customer gives at least 30 days' notice before the initial term ends. The proposal states the amount and payment schedule.
Legal Signals
What each signal meansA 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.
Client Data in Training
Can material a lawyer puts into this product be used to train a model?
The published agreement expressly reserves a right to train on customer content, with no opt out located. Any de identification, anonymization or aggregation qualifier is recorded in the summary.
Section 9.2 gives Deep Legal and its partners the right to use information entered by the customer to improve product functions and prepare general statistics without identifying the customer. The personal data addendum includes an improvement purpose and calls for anonymization where possible. The license does not exclude model training from that improvement use.
Prompt and Output Retention
How long does the product keep what a lawyer typed, and can that be set to zero?
No located public material states how long prompts and outputs are retained.
Daily backups are kept for seven days, access data is deleted within 45 days of termination, and remaining data may be removed after 60. Prompts to the report agent and its answers have no period of their own and no setting a customer can change.
Ethical Walls and Matter Segregation
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.
Each user holds a personal, nontransferable credential, with Active Directory sign in as an option. Portfolios and permissions live in the customer's workspace, and the organization decides which case data feeds its dashboards. Benchmarks of other companies draw only on their public court records.
Third Party Request and Subpoena Notice
If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?
Terms commit to notice where lawfully permitted. No transparency report located.
Section 9.3 requires a party legally obliged to disclose information to notify the other party in a reasonable time. The clause limits disclosure to the confidential information required by law.
Primary Law Corpus Provenance
Where does the law in this product come from, and does the vendor have the right to use it?
Coverage is described by jurisdiction with no identification of the underlying corpus.
Quantum collects Brazilian court cases, appeal decisions and official court journals across three instances, more than one million cases in all, plus CARF tax appeals. Physical, criminal, administrative and sealed cases, matters outside the CNJ numbering standard and cases before 2008 fall outside standard collection. Deep Legal names no license behind the court data or the generative reporting model.
Good Law Verification
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.
Legal Ops reports each case's current procedural stage and closure data from the courts. The analytics count outcomes rather than weigh authority, so no tool shows how a later decision treated a precedent or whether it is still good law.
Refusal and Uncertainty Behavior
What does the product do when the answer is not in the corpus?
No located public material addresses what the product does when it cannot ground an answer.
Forecasts arrive as probabilities and ranges, and Deep Legal warns that ranges widen for small or mixed portfolios. The report agent has no described way to decline a question the case data cannot support, and its answers carry no confidence level.
Fabricated Citation Record
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 has no entry naming Deep Legal or Quantum. The product measures litigation outcomes and forecasts exposure rather than drafting briefs with case citations.
Bar Guidance Alignment
Has the vendor engaged in public with the ethics opinions its buyers are bound by?
Public materials refer to professional responsibility in general terms without naming guidance.
The company's AI guidance discusses lawyer supervision and professional secrecy when analytics or research outputs are used, and classes risk by workflow. It ties Quantum's features to no named OAB or bar opinion, ABA Formal Opinion 512 included.
Billing and Fee Posture
Does the vendor address what happens to the bill when the work takes an hour instead of six?
The product does not touch a fee between a lawyer and a client. It operates before an engagement exists, or it is bought by a team that bills no client for the work. Savings claims aimed at the buyer’s own cost are recorded in the summary and do not make the row a savings claim, because no client bill is in the loop.
Legal Ops finds closed cases still open in the buyer's ERP, so outside counsel fees on them can stop, and the analytics measure the cost of the legal operation itself. Its named customers are companies managing their own litigation portfolios. No feature turns AI assisted work into a line on a lawyer's bill to a client.
Outside Counsel Guideline Readiness
Can a firm get this vendor through a client’s AI clause without a bespoke negotiation?
A current subprocessor or model provider list is published.
AWS and DigitalOcean host the platform. Deep Legal's data addendum offers a route to ask about personal data protection, limits subprocessors and allows a remote audit. The providers of the generative dashboard and report agent go unnamed, and no security certification covers the service.
Court Disclosure Support
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.
The workspace keeps court decisions, case data and results, exports selected data to Excel and turns reports into presentations. Agent answers and signals carry evidence from individual cases and court excerpts. No single export joins the model used, the sources retrieved and a lawyer's verification for a filing.