Bloomberg Law vs Blue J: how they compare in 2026
Bloomberg Law and Blue J are a generalist and a specialist, and the specialist publishes considerably more about itself. Blue J does tax only, in three countries. Its security page names thirty one subprocessors with locations, including the AI layer, and states that all data is stored and processed in a single United States region. It says plainly that no other jurisdiction and no on premise option is available. It names OpenAI and Google as its model providers, with signed agreements barring them from training on its data and abuse monitoring switched off at both. It also publishes 1,498 dollars per user per year, with a purchase path that completes online. Bloomberg Law answers on content and on grounding: 15.5 million court opinions, roughly 200 million dockets and more than 5 million statutes and regulations. Its own machine learning citator reports whether a holding is still valid law, and it documents an abstention where no content covers the topic.
At a glance
All 15 axes, side by side
The same grid applied to every vendor in the index, graded from public sources. 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.
The vendor's approach page lists differentiated content as one of four pillars and says the AI does not generate answers on its own but searches a large collection of vetted information. Without the models, 15.5 million court opinions, 200 million dockets, 5 million statutes and regulations, 8,200 practical guidance documents and a newsroom remain, which is what Bloomberg has sold since 2009. Some tools would not survive: BCITE, Docket Key, Points of Law and Litigation Analytics are built with machine learning and do not exist without it. The subscription is flat and all inclusive, and buyers do not buy or meter the AI separately.
Blue J says it does not build or train its own generative models and uses large language technology from OpenAI and Google. What it owns is the curated tax database: full US federal and state coverage, case law and primary and secondary sources, and licensed Tax Notes and IBFD material, updated daily. Without the AI, a licensed tax content collection remains, the kind of product buyers already get from Checkpoint and CCH.
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.
The vendor names retrieval augmented generation with a proprietary guardrail service, says output is grounded in its own expert or primary content rather than the internet, and says Bloomberg Law Answers cites and links the authorities behind each answer, with one click verification. BCITE, the platform's own citator, reports whether a holding is still good law. The system abstains when no content covers a topic or a question is outside legal scope. No result, test set, sample size or error rate from the described benchmarking by former attorneys is published, although the same page commits to giving customers clear documentation of that benchmarking and validation. The vendor says plainly that AI answers are never fully trustworthy and carry a hallucination risk, and tells customers to check them against primary sources.
Every answer carries inline citations and a source list, the relevant section of each source is highlighted, a user can ask for more sources, and a user can question a single source on its own. The corpus is named: primary tax authority plus licensed Tax Notes and IBFD content, updated daily. No accuracy figure, test set or evaluation result appears on any Blue J page. A 90 percent accuracy claim in third party directories refers to Blue J's earlier outcome prediction product and appears on no Blue J page. ValsAI is listed for benchmark testing, and no results are published.
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.
The vendor says it wants its AI to make the job easier, not do it for the user, and describes building one task at a time rather than producing finished work product. Answers carry inline citations with one click verification, abstention is documented with two named triggers, all generative tools are beta tested before release, and customers are told to check answers against primary sources and use their own judgment. Nothing on the approach to AI page, the AI product page or in either agreement describes what happens after an answer proves wrong; no correction or error reporting route is documented, and no confidence threshold or setting is offered. Deep Thinking, which the vendor says automatically plans and runs multistep research before synthesizing results, has no described checkpoint between request and finished answer.
Section 7 of the Subscriber Agreement makes users responsible for judging whether output is suitable and for consulting their own legal, tax and accounting advisors before using it. Section 12 of the Terms of Use calls the platform an information tool only. The privacy notice says no automated decision making of the GDPR Article 22 kind takes place. In the product, users get inline citations, source lists, highlighted passages and questions to a single source. Nothing describes what the system does on its own: no modes, thresholds or escalation, and nothing on what happens after a wrong answer.
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.
The pricing page has three quotes. One names an organization and an outcome: a corporate counsel at Synovos, Inc. says the platform let the team pull state summaries of laws and regulations without calling outside counsel, cutting tens of calls on minor issues and lowering annual legal fees, though with no figure, timeframe or date. The other two are attributed only by role and segment, a small law practice area leader and a large law partner, and both concern the flat pricing rather than outcomes. The home, AI product, approach to AI and pricing pages and the navigation carry no case study with a method, customer count, named law firm or quantified result.
Blue J names about twenty customer firms, most with a named person and title, including Crowe, RSM US, Adams Brown, GreerWalker, Barnes Dennig, Perelson Weiner and the National Association of Tax Professionals. It publishes figures too: three hours saved per user per week, 75 percent less research time, more than 70 percent of users logging in weekly, and one named partner finishing four to five hours of work in fifteen minutes. The percentages are averages across users rather than measured results at a named firm, and no dates appear.
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.
The privacy policy says collected information includes user content, usage, browsing and search history, reading habits and inferences used to profile a user's preferences and behavior. It describes advertising over that material: lookalike or matched audiences on social platforms, sharing a hashed email with ad companies, and sharing for cross context behavioral advertising under California and Virginia law, with opt outs. The policy covers the marketing site and the products together and does not say whether in product search history is excluded. The subscription terms say the site is not meant to hold personal or health information, give no warranty on how it would be held, and reserve a right to audit and monitor use. The documents do commit to notice of subpoenas, Standard Contractual Clauses, an intragroup data protection agreement, and data centers used only for the vendor's operations. No statement on training with customer input, retention period, separation between users or matters, or treatment of privilege or work product is published.
Blue J is opted out of abuse monitoring at OpenAI and Google, holds signed agreements barring both from training on Blue J data, and says no customer files are used for training. Uploaded files older than 24 hours are deleted automatically, and any present or past client can have all data erased on written request. The Terms of Use carry a mutual confidentiality article at section 10. Nothing describes separation between customers, users or matters, and nothing addresses privilege or work product beyond saying no solicitor client relationship arises with Blue J. The contract also qualifies the training position: section 3 of the Terms of Use licenses what users enter for purposes that include improving the platform.
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.
Both the website terms and the subscription terms say in capitals that none of the services is legal advice, opinion or recommendation, that users needing legal help should consult a competent professional, that users take full responsibility for decisions and conclusions reached through the services, and that no attorney client relationship forms with the vendor or any content supplier. Eligible users are defined: eighteen or older, members of a professional, business or academic community doing professional research, working in a court, government agency, corporation or similar body, or law or business students, with bots excluded. The public approach to AI page addresses competence and supervision, telling customers to check answers against primary sources and use their own judgment, and saying the AI makes the job easier rather than doing it. No bar or ethics guidance is named, including ABA Formal Opinion 512, and no jurisdictional limit on answers is stated.
Section 7 of the Subscriber Agreement says Blue J does not give legal, tax, accounting or other professional advice, that the product is not a substitute for a lawyer's or accountant's judgment, and that users must check output and consult their own advisors. Section 12 of the Terms of Use repeats it and says no solicitor client relationship is formed. No jurisdiction limits for advice are stated, and no professional guidance is named, neither ABA Formal Opinion 512 nor AICPA standards.
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.
The approach to AI page, last modified 18 June 2026, sets out four pillars and gives the accuracy pillar substance: retrieval augmented generation plus a proprietary guardrail service, output grounded in vetted first party or primary material rather than the open internet, beta testing of every generative tool before release, and accuracy evaluation by former attorneys from development through internal, beta and customer testing and over the product's life, with continuous tuning against model drift. It commits to documenting where and when products use generative AI, the sources used, and the benchmarking and validation process. No accountable owner is named, and no test result is published against that benchmarking commitment. Nothing addresses bias or uneven output across matter types, jurisdictions or populations, and the platform profiles named judges, attorneys and firms. Its trust center lists a document titled AI Frequently Asked Questions.
CTO Brett Janssen is named as owner of security compliance. Fifteen security policies are listed and reviewed yearly, risk assessments run yearly on the Cloud Security Alliance CAIQ, and Drata monitors more than a hundred controls. The subprocessor table lists ValsAI for benchmark testing of AI responses. No AI governance framework, owner of model behavior, description of what is tested before a change ships or what the benchmark measures, or ISO 42001 certification is published, and nothing addresses uneven output across taxpayer types or entity structures.
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.
The privacy policy, updated 9 July 2026, says personal information is held on secure servers under administrative, technical, personnel and physical measures, and sets retention by criteria rather than a number: the time needed for the purpose, whether the user has stopped using the products, any legal proceedings, and legal or regulatory requirements. It describes a personal data governance program, a data map of flows within the vendor's systems and to third parties, ongoing testing of safeguards, company wide privacy training, and data centers used only for the vendor's own products and operations with monitored access. Standard Contractual Clauses and an intragroup data protection agreement cover transfers, with further processing terms available where the vendor acts as processor. The trust center adds data erasure, backups, access monitoring, least privilege and logging, and a full responsible disclosure policy. No retention period, named subprocessor or breach notification commitment is published.
The subprocessor table names thirty one entities with role and location, including OpenAI, Google and Microsoft for AI services, Pinecone for vector search, Elastic Cloud, Exa Labs for web search and ValsAI for benchmark testing. Uploaded files older than 24 hours are deleted automatically, and any present or past client can have all data erased on request, with confirmation. Access control includes enforced two factor authentication, least privilege, annual access reviews, Auth0, and BastionZero with SSH disabled. An incident response plan is tested yearly, with prompt reporting to required parties. Data is encrypted with AES-256 at rest and at least TLS 1.2 in transit, backed up daily, and tested by annual penetration tests and quarterly scans. No fixed breach notice deadline to customers 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.
The subscription terms, updated November 2025, cap total liability at 500 dollars whatever the form of action, disclaim all warranties including merchantability, fitness for purpose and noninfringement, provide the services as is and as available, and do not guarantee accuracy, correctness, completeness or timeliness. One exclusion covers research itself: no liability for any inability or failure to perform legal or other research, or to do it properly or completely, even with the vendor's help. Users are solely responsible for the accuracy and adequacy of data used and the output, and any action must be brought within one year. Virginia law governs, with exclusive jurisdiction in Arlington County. The only exception to the cap is death or personal injury from negligence where the law forbids limiting it. Nothing runs to the customer: no indemnity, output warranty or AI specific term. Its trust center lists a certificate of liability.
Section 13 of the Terms of Use disclaims all warranties, including accuracy and completeness, excludes damages of any kind, and caps total liability at ten US dollars however many claims are made. On a $1,498 annual plan, that is under one percent of one user's fee. Section 14 has the customer indemnify Blue J, including for use in legal activities, and no indemnity runs the other way. The Tax Analysts terms in section 13 commit Tax Analysts to defend reasonable claims that the licensed content infringes someone's rights. No remedy applies when an answer is wrong.
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.
Two integrations are described. The vendor's newsroom describes Bloomberg Law Contract Solutions as integrating with Microsoft Word alongside a central contract repository, and the subscription terms cover users reaching Bloomberg Law through the Bloomberg Professional service, the Terminal, whose own agreement then applies. The navigation, the AI product page, the approach to AI page and the footer show no integrations page, developer or API documentation, or named connection to document management, matter management, ebilling or court filing. Dashboard Legal is a project management and collaboration product, not a bridge into a firm's systems.
No integration is published, and none is claimed. The home page, how it works, the pricing page and feature matrix, the security page, the Subscriber Agreement, the Terms of Use and the privacy notice show no integrations page, connector, API or developer documentation, and no mention of document, practice, tax preparation or workflow software. Answers export to PDF and Word, and team plans offer single sign on, which are export and login rather than integration. A 2023 announcement described planned API integrations, and nothing current shows they shipped.
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.
The privacy policy's intragroup data protection statement says the vendor's data centers serve only its own products, services and operations, with secure, monitored access. Transfers are acknowledged: personal information may be processed in various countries with differing protection, including the United States, under Standard Contractual Clauses and an intragroup agreement naming exporters and importers. Nothing published offers a region or residency choice, describes the platform's tenancy, separates processing from storage, or says where generative features run. Its trust center lists a data center item. These statements sit in a personal data annex, not in product or security material.
All data is stored and processed in the AWS us-east-1 region in North Virginia. Blue J states it does not currently offer storage in any other jurisdiction, or an on premises option. Hosting is named as Cloudflare and AWS, every listed subprocessor is in the United States, and the privacy notice covers transfers of European and UK personal data. Nothing says whether the platform is multi tenant or how one customer's data is kept apart from another's.
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.
Every page footer links to the trust center under Trust and Security, with a product identifier in the address. It is a SafeBase portal reachable without a sales call, listing a penetration test report, security whitepaper, SOC 1 and SOC 2, a SIG Lite self assessment, a certificate of liability, a business continuity management document and an AI FAQ, along with control areas for product, access, application and network security, business continuity and training, and a full responsible disclosure policy. SOC 1 and SOC 2 are named without a Type, and no auditor, coverage period, report date or scope is published. Documents are split into public and private, with an access request whose terms are not stated.
An independent auditor maintains a SOC 2 Type 2 report annually, Drata monitors more than a hundred controls, penetration testing runs yearly and vulnerability scans quarterly. A trust center is linked by full URL. Blue J states it provides the SOC 2 report to customers after a signed nondisclosure agreement, so a prospect cannot reach it freely. No auditor, report date, coverage period or trust services criteria are published.
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 vendor publishes retrieval augmented generation as the framework, a proprietary guardrail service over it, generation grounded in its own expert and primary content rather than the internet, and large language models trained on that content. It also cites machine learning, natural language processing, information retrieval, recommendation systems and large language models, and says Bloomberg Industry Group employs hundreds of AI researchers and engineers who publish research each year. Nothing published says whether a third party foundation model sits underneath: the phrase used is our large language models, without saying they are built in house. No model name, version, inference location or subprocessor is given, and there is no commitment to notify customers of model or training changes. Its trust center lists a subprocessors item and an AI FAQ.
Generative models come from OpenAI and Google, since Blue J does not build or train its own. The subprocessor table adds Microsoft for cloud AI, Pinecone for vector search, Elastic Cloud for search and Exa Labs for web search, all in the United States, with the platform on AWS us-east-1. Abuse monitoring is switched off at OpenAI and Google, and signed agreements bar both from training on Blue J data. No specific model is named, so which model answers, and when it changes, is not stated, and no notice of model or provider changes is promised.
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 charging model is stated: one platform, one flat all inclusive price, with unlimited access rather than per search or per document charges. A customer quote on its pricing page says occasional use of a tool costs nothing extra, and the request form names two products, Bloomberg Law Research and Bloomberg Law Dashboard Legal. The pricing page, getting started entry, AI product page and navigation give no figure, unit of charge, tier names, feature splits, term length or implementation charge. Every route to a number is a form, and the Request Pricing page is a lead capture form. Its firm size ranges, from solo to fifty or more practitioners, suggest price varies with headcount, but they are form fields, not a published price structure.
The individual plan is published at 1,498 US dollars per user per year, with what it includes, a seven day free trial with no credit card, and an online purchase path. A feature matrix of about twenty five rows shows what separates sole practitioner and team access, including support, analytics and single sign on. The page promises no add ons and no extra fees, and plans are annual, changeable at any time, with access kept to the end of a canceled period. The team price, which is what most firms buy, is contact for pricing, and enterprise plans go through sales. Canada and United Kingdom pricing pages exist; the figure is the US one.
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.
Four buyer groups have solution pages: law firms, in house counsel, government and law schools, and the government page promises unlimited access to state and federal coverage. The vendor's forms list law firms, tax and accounting firms, corporations, nonprofits and associations, consulting firms, education institutions and government, with firm size ranges from solo to fifty or more practitioners. Content is quantified precisely: more than 15.5 million court opinions growing by about 50,000 a month, about 200 million dockets with more than 21 million pleadings, more than 5 million statutes and regulations, 75 million EDGAR filings and more than 8,200 practical guidance documents, with Docket Key covering over 200 motion and brief types across all federal district courts and Chart Builders comparing rules state by state. Nothing says where coverage stops: no supported practice area list, no coverage table by jurisdiction, no statement of state court depth, and no acknowledged gap.
The practice area is tax only. Four firm sizes have their own pages, from sole practitioner to national, and four use cases are described: advisory, compliance, tax writing and training. The United States, Canada and the United Kingdom each have a separate site with their own pricing, security and contracts, and the US product covers federal, state and local tax. Most named customers are accounting and tax firms rather than law firms. Nothing addresses in house tax departments or government use, and nothing states what the tax content does not cover.
The 12 legal signals, side by side
Recorded rather than graded. These are the questions a practitioner has to answer before a tool touches a client matter, and the answers are taken from public material only.
Client Data in Training
Can material a lawyer puts into this product be used to train a model?
No published term or policy states whether customer input is used to train or refine any model. The privacy policy lists improving, upgrading or enhancing the products and developing new ones among its purposes for collected information, which covers the ground without naming training. The approach to AI page says the vendor's large language models train on its own specialized content, meaning its court opinions, guidance and news, and says nothing about customer material.
The subscription terms prohibit a customer from using product content in prompts or for training, tuning or incorporating into any language model without the vendor's written consent, and define language model broadly enough to catch any generative system. The prohibition runs one way.
Section 3 of the Terms of Use takes a perpetual, irrevocable license over everything a user enters, including results and derivative works, and one of its four purposes is to improve the Blue J Platform. Section 4 of the Subscriber Agreement also lets Blue J review all inputs and outputs to improve functionality. Neither names training, but both allow use of customer content to improve the product, with no deidentification or aggregation limit.
Two points run the other way. The security page says customer data is not used to train generative AI models, that Blue J trains no models of its own, and that OpenAI and Google are barred by signed agreements from training on Blue J data. And section 3 ends with a deletion right on request, beside the perpetual, irrevocable license. Trial users get a narrower, revocable review permission; paying subscribers do not.
Prompt and Output Retention
How long does the product keep what a lawyer typed, and can that be set to zero?
The privacy policy states that personal information is kept for as long as necessary for the described purposes and legitimate business reasons, and names four factors used to set the window: the time needed to achieve the collection purpose, whether the user has stopped using the products, the existence of relevant legal proceedings, and legal or regulatory requirements. Search history and user content are expressly within the categories collected, so questions put to the AI features fall inside this framework, but no period attaches to them and no customer configurable or zero retention setting is offered.
The subscription terms point the other way on uploads, stating the vendor has no obligation to retain or store user generated content.
Uploaded files are deleted automatically once older than 24 hours. Prompts and answers persist as Threads, a product feature, with no published retention period. Deletion is available on request through a customer success manager or the security address, and any present or past client can have all data erased with confirmation. The privacy notice says personal data is kept as long as necessary and may survive termination for business purposes.
Ethical Walls and Matter Segregation
Does retrieval respect the firm’s ethical walls, or can the model read across them?
The privacy policy says personal information may be disclosed to whoever authorized the user's access, such as an employer or school, so a subscribing firm can receive information about what its own users did. The subscription terms treat each registration as single user and require credentials to be kept confidential, which is account security rather than a permission model. Nothing in the privacy policy, either agreement, the trust center landing page or the AI product and approach to AI pages addresses separation between customer organizations, access within a firm, or matters.
Nothing on the home page, how it works, the pricing page and feature matrix or the security page, or in the Subscriber Agreement, Terms of Use or privacy notice, describes separation between customers, between users inside a customer, or between matters, though the security page covers encryption, access and subprocessors at length. Least privilege and two factor authentication govern Blue J's own staff. Workspaces are mentioned as a collaboration unit without saying what they separate, and team plans list custom access controls without explaining them.
Third Party Request and Subpoena Notice
If someone subpoenas the vendor for a firm’s data, does the firm hear about it first?
The privacy policy covers disclosure to unaffiliated third parties under a subpoena, court order or other legal process, and says the vendor will use reasonable efforts to give the user prompt written notice so the user can comment, object and protect the information's confidentiality, and will cooperate at the user's expense. Three qualifications apply: it is reasonable efforts rather than an unconditional promise; cooperation is at the user's cost; and the subscription terms separately reserve the right to use and disclose user information in any legal process or as law requires, with no notice, which nothing reconciles. No transparency report is published.
The privacy notice says Blue J and its providers may disclose personal information in response to a warrant, order or lawful access request from governments, courts or law enforcement. Where law or a court order requires disclosure of a customer's information, Blue J will promptly notify the customer before complying, unless prohibited, and will cooperate on the response. No transparency report is published.
Primary Law Corpus Provenance
Where does the law in this product come from, and does the vendor have the right to use it?
Bloomberg Law publishes counts: more than 15.5 million court opinions growing by about 50,000 a month, about 200 million dockets with more than 21 million pleadings, more than 5 million statutes and regulations, 75 million EDGAR filings and more than 8,200 practitioner guidance documents, covering primary law and the vendor's own editorial layer, with an update rate for opinions. No upstream supplier is named. The privacy policy mentions third party data suppliers only as a category and notes that databases available to registered users contain public records, publicly available information and nonpublic information from those suppliers.
The subscription terms mention third party content providers without naming any. No license, public domain basis or rights statement is published.
Blue J names primary tax authority plus Tax Notes and IBFD, broken down into case law and primary and secondary sources curated by tax experts. Section 13 of the Terms of Use reproduces Tax Analysts' own license terms, including its commitment to defend infringement claims. The database is described as updated daily. No licensing statement is published for IBFD, and no coverage dates or corpus size are published.
Good Law Verification
Does the product tell you when the authority it just cited has been overruled?
BCITE is described as Bloomberg Law's court opinion citator, using machine learning and natural language processing to show quickly whether an opinion's holding is still good law. It is the vendor's own citator with a stated method, not a licensed one or a prompt to check elsewhere. Not disclosed: coverage of courts or dates, an accuracy or error rate, how treatment is decided beyond the techniques named, and whether generative features show BCITE status beside the authorities they cite.
Nothing on the home page, how it works, the pricing page and feature matrix, or in the Terms of Use says whether a returned authority is still good law. The database is described as current and updated daily, which speaks to the collection, not to superseded regulations, revoked rulings or decisions reversed on appeal. The product returns case law, Treasury regulations and rulings as the basis for answers.
Refusal and Uncertainty Behavior
What does the product do when the answer is not in the corpus?
The vendor says Bloomberg Law Answers and AI Assistant will not answer when no content covers the topic or the question is outside legal scope. It also says plainly that AI answers are never fully trustworthy and carry a hallucination risk, and tells customers to check them against primary sources. No worked example of a refusal, measurement of how often abstention works or fails, or confidence score shown to users is published.
None of the home page, how it works, the pricing page and feature matrix, the security page, the Subscriber Agreement or the Terms of Use describes what the product does when it cannot ground an answer, or shows a confidence indicator. An August 2023 announcement quotes Blue J's head of legal research saying the platform is trained to admit when it cannot form an answer supported by authoritative sources. Nothing on the site today repeats that. The current pages offer tools to check answers instead.
Fabricated Citation Record
Does a public court record exist addressing fabricated or hallucinated legal citations in output from this product?
The AI Hallucination Cases database maintained by Damien Charlotin, which tracks court decisions on hallucinated AI content and names the tool where known, records no case naming Bloomberg Law or its publisher, and no court order, opinion or disciplinary record naming the product appears in 2026 sanctions trackers, law library commentary or trade press. This is a statement about the public record rather than a finding about the product.
The name appears often in this literature for other reasons: practitioner guidance names Bloomberg Law as a database for checking citations, and its newsroom reports on sanctions.
The AI Hallucination Cases database maintained by Damien Charlotin, and reporting drawing on it, record no case naming Blue J, Blue J Legal Inc. or BJL US Inc., and no court order, opinion or disciplinary record naming the product is on the public record. This is a statement about the public record rather than a finding about the product. Tax work reaches the Tax Court and revenue authorities more than general courts, and most buyers are accounting firms.
Bar Guidance Alignment
Has the vendor engaged in public with the ethics opinions its buyers are bound by?
The approach to AI, AI product, about and pricing pages, both agreements, the privacy policy and the trust center landing page engage no bar or ethics authority, including ABA Formal Opinion 512 or any state bar guidance, and nothing maps the product to a conduct duty. The disclaimers in both agreements and the validation advice on the AI page state the vendor's own position rather than the rules its buyers follow. Bloomberg Law's newsroom covers legal ethics and AI sanctions extensively; that coverage is journalism, not a statement about the product.
The home page, how it works, the pricing page, the security page, the Subscriber Agreement, the Terms of Use and the privacy notice name no professional guidance: no ABA Formal Opinion 512, state bar opinion, or AICPA or state accountancy board guidance. Most named customers are CPA firms bound by accountancy guidance. Blue J covers the duty in its own terms, at section 7 of the Subscriber Agreement and section 12 of the Terms of Use.
Billing and Fee Posture
Does the vendor address what happens to the bill when the work takes an hour instead of six?
The vendor's material centers on time and cost saved: AI that helps customers work efficiently and finish research faster, summaries to speed reading, and a flat subscription pitched on the basis that occasional tool use costs nothing extra. A customer quote describes lower annual legal fees from fewer calls to outside counsel. The pricing, AI product and approach to AI pages and both agreements describe no record of AI assisted work by matter for fee purposes and no guidance on billing or client disclosure where an AI answer informs advice.
Flat pricing removes the pass through question that metered research platforms raise; it does not address client disclosure.
Blue J claims three hours saved per user per week and 75 percent less research time. Its pricing page calculator turns those into money, with an example of 208 hours, $412,000 of billable capacity and eighteen more client matters a year, based on average savings seen among users. Nothing says how AI assisted work should be recorded, billed or disclosed to clients, and no record of AI use by matter is described.
Outside Counsel Guideline Readiness
Can a firm get this vendor through a client’s AI clause without a bespoke negotiation?
The trust center portal, linked from every product page footer and reachable without a sales call, lists subprocessors, a data processing agreement, data subject requests, a security whitepaper, a SIG Lite self assessment and an AI FAQ. No subprocessor or model provider is named on the portal landing page or any Bloomberg Law page, and documents are split into public and private with an access request whose terms are not stated. Whether a firm can get the listed material without a signed agreement is not stated.
Blue J publishes its subprocessor list openly on the security page: thirty one named entities with role and location, including OpenAI, Google and Microsoft for AI, Pinecone for vector search and ValsAI for benchmark testing. Section 11 of the Terms of Use ties the list into the contract. The SOC 2 report needs a signed nondisclosure agreement, and no data processing agreement is published.
Court Disclosure Support
If a judge’s standing order requires an AI disclosure, can the product produce one?
The vendor commits that generative answers or suggested text include links to sources or an explanation of the result, that Bloomberg Law Answers cites and links supporting authorities, and that inline citations allow one click verification, so a lawyer can produce and check the authorities behind an answer. It also commits to documenting where and when products use generative AI. No model is identified or versioned, so which system produced a passage cannot be known; nothing records who verified an answer; and no export for AI use disclosure or certification is published.
Nothing on the home page, how it works, the pricing page and feature matrix or the security page, or in the Subscriber Agreement or Terms of Use, addresses standing orders, AI use disclosure or certification that a person checked the citations. Answers carry citations and can be exported to PDF or Word, but the export does not record which model produced a passage, what was retrieved or who reviewed it. In tax practice such work reaches revenue authorities more often than courts, and neither setting is addressed.
The questions both sides leave open
Derived from the records above rather than written, so it cannot favor either vendor. Take these into both conversations and ask each side the same question.
- Ethical Walls and Matter Segregation
- Bar Guidance Alignment
Which one fits
Choose Bloomberg Law if
- You need the whole record rather than one field of law. Bloomberg Law quantifies its corpus: more than 15.5 million court opinions growing by around 50,000 a month, and roughly 200 million dockets with over 21 million associated pleadings. It adds more than 5 million codified statutes and regulations and over 8,200 practitioner written guidance documents. Docket Key classifies more than 200 motion and brief types across every federal district court.
- You want the system to tell you when it has nothing. Bloomberg Law publishes retrieval augmented generation alongside a proprietary guardrail service, and states that answers are grounded in its own vetted content rather than the internet at large. Answers carry inline citations with one click verification. It documents an abstention behavior: the tool declines to answer if no content covers the topic or the query falls outside legal scope.
- You want to check whether a holding still stands. BCITE is Bloomberg Law's own citator, built with machine learning and natural language processing to indicate whether a holding remains valid law. It sits inside a flat all inclusive subscription rather than being metered per search or per document, so occasional use of a tool carries no additional charge.
Choose Blue J if
- Your security reviewer wants the whole chain named. Blue J publishes a thirty one entry subprocessor table with locations covering the AI layer itself. It names OpenAI and Google for large language technology, Microsoft for cloud AI, Pinecone for vector search, Elastic Cloud and Exa Labs. Uploaded files are deleted automatically after 24 hours, and any client past or present can exercise a right to be forgotten, with confirmation on completion. Two factor authentication is enforced and the incident response plan is tested annually.
- You need to know where the data sits and what is not on offer. Blue J states that all data is stored and processed in one United States region. It states plainly that it does not currently offer storage in any other jurisdiction and does not offer an on premise option. It names its model providers and states that it holds signed agreements prohibiting them from training on Blue J data, with abuse monitoring switched off at both.
- You want to buy a seat without a sales call. Blue J publishes 1,498 US dollars per user per year, with a seven day free trial requiring no card and a purchase path that completes online. A feature matrix of around twenty five rows separates sole practitioner from team access. Answers carry inline citations, a source list and the relevant passage of each source highlighted, and questions can be put to a single source in isolation.
In summary
Bloomberg Law
Bloomberg Law is a legal research and intelligence platform combining primary law, dockets, practitioner written guidance and a newsroom under one flat subscription. Machine learning runs through the corpus rather than beside it: BCITE as its own citator, Docket Key classifying more than 200 motion and brief types, and litigation analytics. A generative layer provides sourced answers and a chat assistant. According to the AI Legal Index, Bloomberg Law publishes a detailed approach to AI page describing retrieval augmented generation with a proprietary guardrail service and a documented abstention behavior. No accuracy result, named model provider or price is published.
Blue J
Blue J is an AI research platform for tax law sold in the United States, Canada and the United Kingdom. It answers plain language questions from a curated database of primary tax authority with licensed Tax Notes and IBFD commentary. Answers carry inline citations with the relevant passage of each source highlighted, and export as memos and client emails. According to the AI Legal Index, Blue J's case rests on data stewardship: it publishes a thirty one entry subprocessor table covering the AI layer, 24 hour deletion of uploads and a right to be forgotten. It names OpenAI and Google as its model providers and publishes its individual seat price. No accuracy result or integration into practice systems is published.
Questions buyers ask
Bloomberg Law vs Blue J: which one do you need?
They are not substitutes. Bloomberg Law is a generalist platform spanning primary law, dockets, guidance and news for any practice. Blue J is tax only, sold in the United States, Canada and the United Kingdom, and its named customers are predominantly accounting and tax practices rather than law firms. The difference is in what each publishes about itself rather than in what each covers: Blue J names its model providers, its hosting region and its seat price, and Bloomberg Law publishes none of the three. From the AI Legal Index, based on each vendor's own published materials as of September 3, 2026. No vendor pays for placement.
What does each cost?
Blue J publishes 1,498 US dollars per user per year for the individual plan, purchasable online with a seven day trial and no add ons, with team pricing on request. Bloomberg Law publishes the shape and not the number: one platform at one flat all inclusive price with unlimited access rather than per search or per document charges, which is real information in a market that meters, and no figure, unit or tier appears anywhere. From the AI Legal Index, based on each vendor's own published materials as of September 3, 2026. No vendor pays for placement.
Who processes your questions?
Blue J answers it: OpenAI and Google supply the large language technology, Microsoft, Pinecone, Elastic Cloud and Exa Labs appear in the subprocessor table, everything is located in the United States, and both model providers are bound by signed agreements not to train on Blue J data. Bloomberg Law describes its architecture as retrieval augmented generation with a proprietary guardrail service and refers throughout to our large language models, without stating whether any third party foundation model sits underneath or naming one. From the AI Legal Index, based on each vendor's own published materials as of September 3, 2026. No vendor pays for placement.
Does either check whether the law is still good?
Bloomberg Law does, through BCITE, its own citator built with machine learning to indicate whether a holding is still valid law. Blue J does not claim a citator. It answers the currency question differently, by stating that its curated database of primary authority and licensed Tax Notes and IBFD commentary is updated daily. From the AI Legal Index, based on each vendor's own published materials as of September 3, 2026. No vendor pays for placement.
What do Bloomberg Law and Blue J both leave unpublished?
Neither publishes an accuracy result, and both describe a testing regime that would produce one. Bloomberg Law describes an accuracy evaluation with former attorneys running from development through beta and into the life of the product, and commits to giving customers documentation of its benchmarking and validation process. No result, test set or error rate is published. Blue J's own subprocessor table discloses a benchmark testing provider, and no results are published either. Neither names a specific model or version, and neither publishes anything on uneven output. From the AI Legal Index, based on each vendor's own published materials as of September 3, 2026. No vendor pays for placement.
Neither vendor paid for inclusion, placement or a grade, and neither reviewed this page before it published. Everything on it comes from public material on the dates shown. How the index grades.
Both publish liability positions that can be read before signing. Bloomberg Law caps aggregate liability at 500 dollars. It excludes liability for any inability or failure to perform legal or other research or related work, or to perform it properly or completely, even where assisted by the vendor. Blue J caps total aggregate liability at ten dollars, against its own published seat price of 1,498 dollars a year, and its indemnity runs from the customer to Blue J. Separately, Bloomberg Law's privacy policy states that the information collected includes user content, search history and browsing history. It describes creating lookalike audiences for targeted advertising, sharing a hashed email address with advertising companies and sharing for cross contextual behavioral advertising, with opt outs offered. The policy covers the marketing site and the products together and does not state whether in product search history is carved out. Neither vendor reviewed this page.