L
LBOX
LBOX is a Korean legal research and AI platform built on a database of South Korean court judgments, statutes and administrative rules, used on the web and by mobile app. Its AI agent, LBOX AI, runs the whole task itself. It reads the question, searches, checks what it finds and writes the answer, with linked citations to the cases and statutes it relied on. Access to the agent is limited to verified legal professionals, recognized courts and public bodies, and staff working under a qualified professional's supervision.
The company states that queries and uploads are deleted once an answer is produced and are not used to train models. Plans run from an individual Standard plan to firm, law school and public sector plans. LBOX Co., Ltd. is based in Seoul and led by its founder Lee Jin.
LBOX, head to head
Side by side on the same 15 capability axes and 12 legal signals: where the two part company, and what neither one publishes.
Alternatives to LBOX
Vendors in the same categories as LBOX whose graded profiles look most alike, with a line on what each product does.
Similar is not the same, so check that each one does the job you need. No vendor pays for placement.
- OOtto Schmidt AnswersOtto Schmidt Answers is the AI research assistant of Verlag Dr.
- AAnnlexAnnlex is a case law research platform for United States court opinions, operated by Annlex Systems LLC and created by founder Doug Hall.
- GGaius-LexGaius-Lex is an AI research and drafting agent for lawyers and tax advisers covering Polish, Czech, Austrian and German law.
- IIberley IAIberley IA is the AI assistant built into Iberley, the Spanish legal information database run by Iberley Información Legal, S.L. of A Coruña.
- LLextel AILextel AI is an AI legal research and drafting platform for Italian lawyers, built on a legal database of more than 8 million documents updated daily.
- NNormo.aiNormo.ai is an AI assistant for Italian professionals who work with the law, sold to lawyers and law firms alongside accountants, labor consultants, tax assistance centers and other advisers.
Ordered by how closely the graded profiles match across the 15 capability axes, inside the same categories. See how each axis is graded in the methodology.
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.
LBOX began as a judgment database, and the AI came later as LBOX AI. The company rebuilt it as an agent that runs every step from the question to the answer, calls this its most fundamental product change, and claims it as Korea's first legal AI agent. It now sells LBOX as a single AI workspace for search, AI answers and drafting. Beside the AI, LBOX offers a Korean case law and statute database with search, projects, folders and links between instances of a case, and the company treats that data as its core asset.
The agent is limited to verified professionals and supervised staff, so some subscribers use the platform without it. LCUBE, a Korean legal language model the company released with KAIST at NeurIPS 2022, is a research artifact, not the shipped product.
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.
LBOX grounds answers in its own legal data and links every cited precedent and statute so a reader can open it. It states this reduces the hallucination risk of general purpose AI. Where an answer draws on a commentary from its LBOX Scholar imprint, the original text opens from the answer, and authors see how often the AI cites them. No accuracy figure, error rate, test set or evaluation is published for the agent, search or drafting, and nothing says how often answers carry support or what the system says when it finds none.
Press coverage reports the company claiming its agent outperformed human bar exam candidates, and that claim is not published with a test set.
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.
LBOX limits who may run the agent. LBOX AI is open to members verified with documents as lawyers, judicial scriveners, labor attorneys, tax accountants, patent attorneys or accountants, and to staff of courts, prosecution offices, police and legal research bodies it recognizes. Other employees may use it only with a qualified professional's prior approval and under their supervision and review, and public bodies use it on the Public plan, which works by IP address.
The company frames this as keeping legal AI use lawful. The agent itself acts broadly, planning and running the whole sequence from question to answer, and drafting. No modes, confidence threshold, stopping or escalation rule, review step between generation and use, or plan for what happens when it is wrong is described. Opening the cited authority is the user's own check.
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.
Users quoted on the pricing page are identified only by role, with competitor names removed. The Google Play listing shows 10,000 plus downloads and a 4.6 rating from 155 reviews. LBOX Scholar names its authors, who are suppliers rather than customers. No named law firm, company, court or public body, case study or outcome figure is published on LBOX's own pages. Press coverage says LBOX serves ten of the largest Korean law firms, companies and judicial institutions, and an aggregator reports about 23,000 lawyers. Neither figure is published by LBOX or tied to a named customer.
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.
Personal information in queries and uploads is not used to train models, LBOX states. Queries and uploads are used only while the answer is generated and then deleted, with a 30 day ceiling at each overseas processor. Documents are visible only to their creator and to Business colleagues given sharing rights, and material is held encrypted in customer dedicated storage. The privacy policy names OpenAI, Anthropic, Google, Amazon and Microsoft as recipients of LBOX AI input, with purpose and retention for each.
Privilege, professional secrecy and work product are not addressed, though verified lawyers upload case documents, and nothing says what happens if that material is demanded. The no training statement covers personal information in queries rather than the content itself.
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.
Who may use the AI agent is stated and enforced with documents. Verified lawyers, judicial scriveners, labor attorneys, tax accountants, patent attorneys and accountants may use it, as may staff of recognized courts, prosecution offices, police and legal research bodies. Other employees may use it only with a qualified professional's prior approval and under their supervision and review, and public bodies use it on the IP based plan.
The company frames this as keeping legal AI use lawful, which engages Korean rules on who may do legal work without naming them. No statement of what the product is and is not appears, no advice disclaimer appears on the home, AI or help center pages, and no jurisdiction limit is stated, though the content is Korean law throughout. The company's consumer lawyer directory is a separate product.
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.
No AI governance position is published. LBOX documents data protection governance in detail, with a named protection officer (the chief executive), a named handler and department, an internal management plan with inspections, a dedicated team, staff training and a grievance route. None of it covers AI. No AI policy, acceptable use position, owner of model behavior, testing before release or evaluation result is published, and bias is not addressed, although the agent chooses which precedents to show.
The company published a benchmark and model with an academic partner in 2022, and no evaluation of the shipped agent is published. The help center and policies say nothing on this.
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.
Retention, deletion, access control and processors are set out in detail, as Korean data protection law requires. Query text and uploads are used only while an answer is generated and then deleted, with a ceiling of 30 days at each overseas processor. Usage logs are kept until a member leaves, and statutory periods are listed. Electronic records are destroyed unrecoverably and paper shredded or burned, and verification documents are destroyed once checked.
Access controls include rights management, an access control system, encryption, security software, inspected access logs and physical controls on server rooms. All processors are listed, domestic and overseas, including those they delegate to, and LBOX commits to publish changes promptly. No breach notice commitment, deadline or incident response description is published.
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.
No published page states who bears the loss when output is wrong. A customer agreement exists in three parts, the integrated terms, the paid service terms and the operating policy, all linked from LBOX's footer and help center. Notices on the help center set refunds if a paid plan is unused within seven days, or thirty days after a switch from Standard to Business or Public. They give no fee reduction if a Business plan loses users mid term, and fee changes come with notice and an objection window.
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.
No integration with other legal systems is published. Search, AI answers and drafting all happen in LBOX's own interface, with projects and folders for matters and sharing within a Business plan. No practice management, document management, billing or CRM system is named, and no public API, developer documentation or connector catalog is published. There is no integrations page in the navigation or help center. The iOS and Android apps are the same product on another device, and the OCR and payment providers are LBOX's own suppliers. How work leaves the platform is not described.
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.
Each overseas recipient of text and files sent to LBOX AI is listed in the privacy policy, by name, country, method, purpose and retention. The recipients are OpenAI, Anthropic, Google, Amazon and Microsoft in the United States, with transfers to Amazon and Microsoft encrypted in transit, and retention capped at 30 days except Google at 180. The policy states that refusing the transfer abroad means the service cannot be used.
Material is held encrypted in customer dedicated storage, and documents are visible only to their creator and to Business colleagues given sharing rights. No data center, region or cloud provider is named for the main platform, nothing separates storage from processing for data in Korea, and no residency or on premises option is offered.
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.
No certification held by LBOX and no trust center are published. LBOX's home page says all data is processed on verified infrastructure with global security certification, which describes the infrastructure provider, not LBOX. No LBOX certificate or audit report under ISO 27001, ISO 27701, SOC 2 or Korea's ISMS-P is published, and no auditor, certificate number, penetration test summary or security page exists. LBOX describes its own controls in detail: administrative, technical and physical measures, an internal management plan with inspections, a dedicated team, training, access management, access control, encryption, security software, inspected access logs and controlled server rooms.
It asserts compliance with the Personal Information Protection Act and names a protection officer. Neither is an independent attestation.
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.
The privacy policy names every provider that receives what users type into LBOX AI. It lists OpenAI LLC, Anthropic PBC, Google, Amazon and Microsoft by legal entity and contact address, each used to generate answers, with Microsoft also handling optical character recognition of uploads. Upstage is the domestic OCR processor. Each has a retention period, 30 days after use for all but Google, at 180. LBOX commits to publish any change of processor or scope without delay, and further delegation needs its consent.
No model or version is named, nothing says which provider handles which task or how requests are routed, and nothing separates LBOX's own retrieval layer from the outside models.
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.
The plans are Standard for individuals, Business and Enterprise for firms and legal departments, Law School for students, and Public for government bodies on IP authentication. Members of bar associations and of the labor attorney and loss adjuster bodies get partnership discounts. The unit is per user per month for Standard, per seat with volume discounts for Business and Enterprise, and institution wide for Public.
LBOX's fee change notice set the Standard rate moving from 29,900 won to a six month rate of 39,900 won and then to 69,900 won a month. That figure comes from a notice, not a current rate card, and Business and Enterprise are quoted. Refund and fee change rules are published in the operating policy.
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.
An enforced eligibility rule defines who LBOX serves. Professionals include lawyers, judicial scriveners, labor attorneys, tax accountants, patent attorneys and accountants, each with a verification route. Institutions include courts, prosecution offices, police and legal research bodies, plus government bodies on the Public plan. Firms and legal departments use Business and Enterprise plans with administrators, students use the Law School plan, and other staff use it only under a qualified professional's supervision.
Content is Korean: court judgments, statutes and administrative rules, with LBOX Scholar commentaries on top. No practice area is named as supported or not, nothing says which courts or years the judgment database covers, and nothing says whether the agent covers the whole corpus. Korea is the only jurisdiction and is never stated as a limit.
5 public documents
The public pages on file for LBOX, 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.
-
lbox.kr5 signals
Client Data in Training, Ethical Walls and Matter Segregation, Good Law Verification and 2 more
Read Sep 13, 2026
-
Prompt and Output Retention, Third Party Request and Subpoena Notice, Outside Counsel Guideline Readiness
Read Sep 13, 2026
-
lbox.kr/scholar1 signal
Primary Law Corpus Provenance
Read Sep 13, 2026
-
lbox.kr/v2/pricing1 signal
Billing and Fee Posture
Read Sep 13, 2026
-
Bar Guidance Alignment
Read Sep 13, 2026
From ₩69,900 per month
- LBOX publishes its plans openly, and the individual price is known, though the current figure may differ.
- There are five plans. Standard is for individuals. Business and Enterprise are for firms and legal departments, sold per seat with a volume discount and an administrator who manages access. Law School is for students, including repeat bar candidates and doctoral students. Public is for courts, prosecution offices, police and other public bodies. It works by IP address rather than individual logins, and the company describes it as a social contribution.
- The company's fee notice set Standard at 69,900 won a month, up from 29,900 won through a six month rate of 39,900 won. That comes from a notice of change rather than a current rate card. Business and Enterprise are quoted.
- Members of bar associations and of the labor attorney and loss adjuster bodies get partnership discounts.
- You get a refund if you have not used a paid plan within seven days of paying, or within thirty days after switching from Standard to Business or Public. If a Business plan loses users mid term, there is no refund for the unused seats. Fee changes are announced in advance with a window to object.
- The contract is published in three parts.
There are five plans. Standard is for individuals. Business and Enterprise are for firms and legal departments, per seat with volume discounts and administrator controls. Law School is for eligible students, including repeat bar candidates and doctoral students, and Public is for government bodies on IP authentication, described as a social contribution. LBOX's own fee change notice set the individual rate at 69,900 won a month, reached from 29,900 won via a six month rate of 39,900 won, with existing subscribers offered the lower rate if they accepted the change within a stated window.
Bar association, labor attorney and loss adjuster members get partnership discounts. AI is not metered separately but limited by eligibility to verified professionals, staff of recognized judicial and law enforcement bodies, and supervised employees, so a Business administrator assigns AI rights. Refunds apply to a plan unused within seven days of payment, or thirty days after a switch from Standard to Business or Public, with no refund when a Business plan's users fall mid term.
Changes are announced in advance with an objection window, and silence counts as acceptance. No term length, minimum or price uplift is stated on the pricing page or in the help center notices.
Confidentiality and data terms: No business associate agreement, data processing addendum or separately signable data instrument is published. The data commitments live in the privacy policy and on the product pages, and Korean data protection law requires much of the detail. They include a full processor list with the parties those processors delegate to, and a table of transfers abroad naming OpenAI, Anthropic, Google, Amazon and Microsoft as recipients of user input to the AI, with purpose and retention for each. Query text and uploaded files are deleted once the answer is generated, each overseas processor has a 30 day ceiling except Google at 180 days, personal information in queries and uploads is not used to train AI models, and any change of processor is to be published without delay. LBOX accepts the statutory roles under the Personal Information Protection Act and names a protection officer, a handler and a responsible department. No certification held by LBOX itself is published under any standard, including Korea's own ISMS-P, no auditor or report period is named, and no trust center exists. The only certification mentioned is that of the underlying infrastructure provider. No breach notification commitment to customers is published.
Note: Plans, discounts and refunds come from LBOX's pricing page copy and the operating policy notices on its help center. The entry price is shown in Korean won, as published, with no conversion. It is the Standard monthly rate from LBOX's own fee change notice, which set 29,900 won moving to a six month 39,900 won and then 69,900 won. The figure comes from a change notice rather than a current rate card. Business and Enterprise are quoted per seat with volume discounts, Public is institution wide by IP, and Law School has defined eligibility. The customer agreement behind the price is published in three parts in the footer.
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?
A public policy or trust page states no training on customer content, with no matching term located in the published agreement.
LBOX states that personal information in queries and uploaded files is not used to train AI models. It also states that queries and uploads are used only while an answer is generated and are deleted once that purpose is met, and the privacy policy caps retention at each overseas model provider at 30 days. The statement covers personal information in queries and files rather than the content itself. The customer agreement is published in three parts in the footer.
Prompt and Output Retention
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.
LBOX publishes fixed retention at two levels. At product level, query text and uploads are used only while an answer is generated and then deleted. At processor level, the privacy policy caps retention at 30 days after use for OpenAI, Anthropic, Amazon and Microsoft. Google's entry says 180 days, with no explanation. Usage data, including search terms, clicks, AI query history and document creation, saving and download history, is kept until a member leaves and is collected through named third party software kits.
Statutory periods are listed, such as three months for communication records and five years for contract and payment records.
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.
A permission model works at document level. Material a user creates is visible only to that user and to Business plan members granted sharing rights, a closed default with explicit grants. Material is held encrypted in customer dedicated storage, and work is organized into projects and folders that carry the sharing boundary. Business and Enterprise plans separate administrators from users, and the help center covers a departing employee's projects.
No list of roles or permission levels is published, and nothing says how grants are made or revoked, whether administrators can read users' projects, or whether access is logged. Conflicts walls are not addressed as such, and nothing limits which LBOX staff may view customer material.
Third Party Request and Subpoena Notice
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 allows LBOX to provide personal data to a third party when a state agency requests it under a procedure set by law. The trigger is tied to a statutory procedure, and other third party sharing is consent based and listed. The same clause commits LBOX to state any such disclosure in the privacy policy itself, which is public notice after the event rather than notice to the customer. Nothing says the subscriber would be told of a demand, given a chance to object, or told afterwards, and no transparency report exists.
Primary Law Corpus Provenance
Where does the law in this product come from, and does the vendor have the right to use it?
Sources are identified without stating the license or rights basis.
The corpus is South Korean court judgments across instances, statutes and administrative rules, described in LBOX's 2022 research as 147,000 precedents and 259 million tokens. Nothing says how judgments are obtained, under what access route or agreement, or what LBOX may do with them, in a country where access to judgments is contested. One part is licensed openly. LBOX Scholar commissions commentaries and practice texts from named academics and practitioners, shows them as citation support in AI answers, and pays royalties based on actual use. That covers secondary material, not the precedents.
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.
A case is linked across instances, from first instance to appeal to the Supreme Court, which users praise, and cited cases and statutes open from AI answers. That linking is case history, not treatment. Nothing shows whether a case's proposition has since been overruled, distinguished or superseded, and no treatment flag exists. Nothing on the home page, the AI landing page, the Scholar pages, the help center including release and policy notices, or the mobile listing addresses later treatment.
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.
Nothing published says what the system does when it cannot answer reliably. LBOX says answers come from its own legal data to reduce hallucination. Nothing says the agent declines questions outside its corpus, reports finding no authority, shows a confidence signal, marks weak passages, or escalates. The agent plans, searches, judges and answers on its own, and LBOX says verification is a step in that loop, without saying what the step does when it fails. Linked citations let a lawyer check the answer.
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.
No decision, order or disciplinary finding in Korean legal press or international trackers names LBOX or LBOX AI, under the company name in Korean or English or the product name. Korean legal press reported in 2026 the first domestic case of fabricated citations reaching a court: five judgments cited by an advocate did not exist, and the advocate admitted using AI. Other reporting describes a court ordering a party to explain a fabricated citation and produce the Supreme Court originals relied on. No product was named in that reporting, and the accounts describe general purpose assistants.
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.
LBOX verifies lawyers and other professionals with documents, admits recognized courts, prosecution offices, police and research bodies, and allows other staff only under a qualified professional's approval and supervision, framing this as keeping legal AI use lawful. The Korean Bar Association appears only as a source of membership numbers for verification and as a discount partner. No provision of the Attorney-at-Law Act, Korean Bar Association opinion, ethics rule or court guidance is cited, though Korean courts have started demanding explanations for nonexistent citations. Nothing connects the agent's output to the verifying duty of the lawyer who files it.
Billing and Fee Posture
Does the vendor address what happens to the bill when the work takes an hour instead of six?
The product sits inside a lawyer to client fee relationship and no located public material addresses billing, fee or disclosure treatment, with no savings claim published either.
Nothing published addresses how AI assisted work is billed. LBOX's buyers are verified lawyers and firms. Research time in Korea is usually either billed or absorbed into a fee, and nothing says which applies. No record per matter of AI assisted work, marking of machine output for billing, or fee or disclosure guidance is described, and no saving is claimed in billable terms. LBOX's per seat pricing, discounts and refund rules do not address the client's bill.
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.
The privacy policy lists every processor, domestic and overseas, including parties they delegate to: payment and identity providers, Upstage for OCR, and messaging suppliers. Its table of transfers abroad names each recipient of LBOX AI input by legal entity, OpenAI LLC, Anthropic PBC, Google, Amazon and Microsoft, with country, transfer method, purpose, contact address and retention of 30 days after use, except Google at 180.
LBOX commits to publish processor changes without delay. No data processing addendum or consent pack is drafted to send to a client. The disclosure sits in a privacy policy written for individuals under Korean law.
Court Disclosure Support
If a judge’s standing order requires an AI disclosure, can the product produce one?
No located public material addresses court disclosure or verification certification.
Nothing published addresses disclosing AI involvement in legal work. Nothing marks output as machine generated once it leaves the platform, and no audit trail, record per matter, export, certification template or disclosure guidance is described. LBOX keeps AI query history as usage data for its own analytics until a member leaves, and nothing says the subscriber can use it as evidence. Linked citations support checking before filing, and project storage is storage, not provenance.
Korean courts have begun ordering parties to explain how a fabricated citation was filed and to produce the originals relied on.