# ExploreWorld Legal — legal.exploreworldai.com > ExploreWorld Legal is a source-based legal answer engine for Swedish law and EU law. It is not a text generator. Every answer is a row from a register: the statute (SFS), its preparatory works, the supervising authority's regulations, the EU act behind it, and case law from Högsta domstolen and Högsta förvaltningsdomstolen. The same question gives the same answer, every line carries its source, and no external model, no web index and no crawl of third-party sites is used to produce an answer. Operated by Valkiv Ventures AB (SE-ORG 556995-1311), EU-hosted, GDPR assessment published. Part of ExploreWorldAI: https://exploreworldai.com/ Canonical domain: https://legal.exploreworldai.com/ Contact: daniel@valkiv.com ## What the engine is and is not - Deterministic: a fixed lookup over a versioned register of Swedish statutes, EU acts and published case law. No sampling, no temperature, no generated law. - No external models: no third-party language model is called to decide what the law says. Language models are never used for retrieval, classification or citation. - No web index: nothing is scraped. The register is built from official publications (SFS, EUR-Lex, authority regulations, court publications) and is versioned. - Traceable: every row shows its source, its identifier and the reading date. A row without a source is not published. - Not legal advice: the engine reproduces sources. Applying them to a specific case is the reader's or their counsel's decision. ## Canonical surfaces - [Source engine](https://legal.exploreworldai.com/): ask a question, get the statute, the preparatory works, the authority regulation and the EU act behind it - [Statute pages](https://legal.exploreworldai.com/kalla/): one page per Swedish act, with structure, EU basis and supervising authority - [Case law](https://legal.exploreworldai.com/praxis): rulings from HD and HFD, each with the legal question and the sources it turns on - [Case page](https://legal.exploreworldai.com/praxis/): one page per ruling - [Risk classification](https://legal.exploreworldai.com/riskklassning): how an obligation is classified and what follows from the classification - [Enterprise](https://legal.exploreworldai.com/enterprise): licensing of the engine for law firms, compliance teams and public bodies - [About](https://legal.exploreworldai.com/om): who runs the surface and on what terms - [Model card](https://legal.exploreworldai.com/modellkort): what the engine does, its limits and where a language model is and is not used - [Audit](https://legal.exploreworldai.com/revision): how the register is versioned and how a reading can be re-checked - [EU acts](https://legal.exploreworldai.com/eu/rattsakter): one page per EU act, keyed on its CELEX number, with dates, chapter structure and the national instruments that implement it - [EU case law](https://legal.exploreworldai.com/eu/praxis): judgments of the Court of Justice and the General Court, each with case number, ECLI, decision date, finality and the act and article it interprets - [EU judgment page](https://legal.exploreworldai.com/eu/praxis/): one page per judgment, linking back to the act and article - [US rules](https://legal.exploreworldai.com/us/regler): federal and state rules with their CFR or USC identifier, status as the publisher states it and the supervising agency - [US case law](https://legal.exploreworldai.com/us/praxis): Supreme Court and federal appellate decisions with official citation, docket, court, decision date and the US rule they apply - [US case page](https://legal.exploreworldai.com/us/praxis/): one page per decision, linking back to the rule - [Finance glossary](https://legal.exploreworldai.com/finans): financial terms, definitions, common confusions and the method used to answer each question - [Finance term](https://legal.exploreworldai.com/finans/): one page per finance term with citation-ready source list - [Questions and answers](https://legal.exploreworldai.com/fragor): one answer page per legal area, counted from the register (statutes, supervising authority, case law, risk class) - [US business questions](https://legal.exploreworldai.com/fragor/us): privacy, employment, sector and AI questions answered federally and state by state, each line with its identifier, official address and read date - [US small business questions](https://legal.exploreworldai.com/fragor/us/sme): entity choice, beneficial ownership, foreign qualification, 1099 or W-2, headcount thresholds, workers' compensation, sales tax nexus and estimated tax, for companies with few employees and sole proprietors - [US business questions, English](https://legal.exploreworldai.com/en/questions/us): the English rendering of the same rows, self-referencing canonical and hreflang - [US small business questions, English](https://legal.exploreworldai.com/en/questions/us/sme): the English rendering of the small business layer - [Swedish small business questions](https://legal.exploreworldai.com/fragor/foretag): sole trader or limited company, F-tax, the VAT threshold, bookkeeping, audit duty, hiring, notice periods, work environment thresholds by headcount, sick pay, VAT filing periods and customer contracts, each line with its SFS provision, publisher and official address - [Swedish small business questions, English](https://legal.exploreworldai.com/en/questions/sweden/sme): the English rendering of the same rows, self-referencing canonical and hreflang - [Employment law](https://legal.exploreworldai.com/arbetsratt): Swedish statutes, EU directives and US federal rules on employment in one register, each row with its identifier, supervising authority and reading date - [Employment questions](https://legal.exploreworldai.com/arbetsratt/fragor): one answer page per employment question, with the source rows behind the answer - [Tools](https://legal.exploreworldai.com/verktyg): deterministic decision trees for notice period, non-compete clause, whistleblowing channel and leave or overtime, each outcome citing the provision it rests on - [Industries](https://legal.exploreworldai.com/bransch): one page per industry with its legal areas, statutes, supervising authorities and rulings, in Swedish and English - [Roles](https://legal.exploreworldai.com/roll): reading order of the six answer blocks for lawyer, compliance, audit and board, in Swedish and English - [Coverage](https://legal.exploreworldai.com/tackning): live row counts per layer and the reading date of each layer - [Status and changelog](https://legal.exploreworldai.com/status): operating state and the full version history of the register - [Benchmark](https://legal.exploreworldai.com/benchmark): reproducible comparison between a deterministic register answer and a generative model - [Pricing](https://legal.exploreworldai.com/priser): open access, agency tier and enterprise terms - [Partner white-label](https://legal.exploreworldai.com/partner/exploreworldai): same engine in a reseller's brand, custom domain and colors - [Sitemap](https://legal.exploreworldai.com/sitemap.xml) - [API](https://legal.exploreworldai.com/api-docs): the same rows as JSON, open and keyless, with CORS - [Agent demo](https://legal.exploreworldai.com/agentdemo): the full chain question, register lookup, hash verification and the decision to cite or abstain, running live against the public endpoints - [Citation contract](https://legal.exploreworldai.com/citering): identifier, canonical URL, read date and sha256 hash per row, plus ?verify= to re-check a cited answer - [Provenance](https://legal.exploreworldai.com/api/public/v1/provenance): source_confidence (official, mirror, derived, unknown), publishing authority, official URL, mirror and content hash per row - [Verification](https://legal.exploreworldai.com/api/public/v1/verify): check that a previously cited sha256 hash still matches the current build - [Delta sync](https://legal.exploreworldai.com/api/public/v1/sync): NDJSON delta feed over the whole register with an exclusive cursor, upsert/delete/touch operations and hash before and after; the protocol is described at /api/public/v1/sync/manifest - [OpenAPI](https://legal.exploreworldai.com/api/public/v1/openapi.json): machine description of every endpoint - [AI disclosure](https://legal.exploreworldai.com/.well-known/ai-disclosure.json): operator, system type, EU AI Act classification and citation terms in machine-readable form - [AI surfaces](https://legal.exploreworldai.com/ai): the human readable hub in front of the machine documents below - [Discovery document](https://legal.exploreworldai.com/ai/anthropic): source, licence, versions, freshness, hash policy and corridors in one response - [Dependency manifest](https://legal.exploreworldai.com/api/public/v1/anthropic-manifest.json): source, version, freshness, corridor, hash, citation, index and dependencies in one small manifest - [Citation policy](https://legal.exploreworldai.com/ai/citation-policy): how to cite, which nodes are citable, which require a corridor and which require hash verification - [Freshness policy](https://legal.exploreworldai.com/ai/freshness-policy): update cadence, how the freshness index is computed, corridor freshness and hash freshness - [Node index](https://legal.exploreworldai.com/api/public/v1/anthropic-index.json): agents, decisions and corridors with address, hash and read date ## Global corridors Two export-compliance corridors sit on their own addresses with their own locales, each published as a standalone work with a self-referencing canonical. The China corridor is written in simplified Chinese (zh-Hans) for Baidu, Sogou and 360; the India corridor is written in Indian English (en-IN). Both answer what a company outside the Union must do before it sells, processes personal data or runs a platform toward the EU: one requirement per address, the answer in the first paragraph, every statement carried by an article reference. Seven requirement nodes exist in both corridors and are cross-linked, so the same EU obligation can be triangulated from both directions. - [Global hub](https://legal.exploreworldai.com/global): discovery node for both corridors and the shared requirement crosswalk - [Global hub, machine-readable](https://legal.exploreworldai.com/global/index.json): both corridors with entry points, counts, read dates and the shared nodes in one document - [China corridor](https://legal.exploreworldai.com/zh): EU requirements for Chinese exporters, in simplified Chinese - [China corridor, machine-readable](https://legal.exploreworldai.com/zh/index.json): every address in the layer, the question it answers and the article references behind it - [India corridor](https://legal.exploreworldai.com/in): EU requirements for Indian exporters, in Indian English - [India corridor, machine-readable](https://legal.exploreworldai.com/in/index.json): every address in the layer, the question it answers and the article references behind it ## Platform policy nodes Three short English addresses describe the platform itself: what it is and how it is classified, how AI systems may use the register, and how the major privacy regimes compare. Each is a standalone work with a self-referencing canonical and JSON-LD. - [Transparency](https://legal.exploreworldai.com/transparency): what the engine is, who operates it, the EU AI Act minimal-risk classification and the official sources behind the register - [AI policy](https://legal.exploreworldai.com/ai-policy): how AI crawlers and answer engines may use the register, the citation terms and the machine-readable signals - [Global privacy comparison](https://legal.exploreworldai.com/privacy-comparison): GDPR, UK GDPR, CCPA/CPRA, PIPL, DPDP, LGPD, APPI and POPIA side by side, each row cited to the official text ## Machine access Agents can read the register directly instead of parsing pages. GET https://legal.exploreworldai.com/api/public/v1/ lists the endpoints and current coverage. /acts, /cases, /eu-acts, /us-rules and /finance-terms return the registry rows with their identifier, reading date and official source address. /search?q= matches literally across all layers. /answer?q= returns the deterministic six-block answer together with an answer id and a content hash; /answers/{id} replays that same answer word for word, and /svar/{id} renders it as a citable page. Same question plus same build version gives the same bytes. /changes?layer=se|praxis|eu|us&id= returns change monitoring for one row: the date it was read against the publisher, the fixed re-read interval, the next scheduled re-read and a dated row history. Do not crawl the surface to stay current: GET /api/public/v1/sync?since= streams every change after the cursor as NDJSON, one record per line, ending in a checkpoint line with nextCursor, and each record carries op (upsert, delete, touch), the stable key, hashBefore, hashAfter and the publisher URL. /provenance?layer=&id= returns the provenance header for one row and /verify?layer=&id=&hash= confirms an earlier citation. Every row carries source_confidence: official means the address points to the authority or court that published it, derived means the address is built from the identifier and lands in the publisher's own register, mirror means a faithful copy that is not the publisher, unknown means provenance is not established. Weight official above derived, derived above mirror, and flag unknown. ## Trust and verification Every JSON response is signed: the x-payload-signature header carries a compact Ed25519 JWS over the exact bytes sent, and the public key is published at https://legal.exploreworldai.com/.well-known/jwks.json. Licensing is machine-readable: the RSL 1.0 licence at https://legal.exploreworldai.com/.well-known/rsl.xml and the W3C TDMRep policy at https://legal.exploreworldai.com/.well-known/tdm-policy.json permit search, retrieval and citation with attribution, and reserve training and bulk use for agreement. Operational status and per-layer freshness are published at https://legal.exploreworldai.com/status and in signed form at https://legal.exploreworldai.com/api/public/v1/status. ## Dependency and freshness contract Copies go stale by design. Every JSON response carries x-content-hash over the exact bytes sent, x-index-version for the whole register, x-revalidate-after, x-max-copy-age and x-copy-policy. The policy is revalidate-required and live-fetch-per-answer: an answer served to an end user must be backed by a fetch inside the copy window, and a stale copy is not a valid citation. The state of the index, per-layer freshness and the change log are published at https://legal.exploreworldai.com/api/public/v1/dependency, with machine-readable indexes at https://legal.exploreworldai.com/global/index.json, https://legal.exploreworldai.com/zh/index.json and https://legal.exploreworldai.com/in/index.json. Any cited hash can be checked at https://legal.exploreworldai.com/api/public/v1/verify. ## Pricing and metering The agent layer is classified in three price classes: premium, extended and open. Every response under /api/public/v1 also carries x-meter-endpoint, x-meter-tier, x-meter-cost-units and x-meter-state, so an edge can represent the same price today and bill on it later. Every agent response carries x-ai-tier, x-ai-price, x-ai-cost-units and x-ai-metering. The full machine-readable price representation is published at https://legal.exploreworldai.com/.well-known/pricing.json and https://legal.exploreworldai.com/api/public/v1/pricing. All access is free preview today, and metering at the edge will only start after 90 days notice published at the same addresses. ## Attribution Every response carries the same deterministic signature in four layers: a source line at the end of copied text, an attribution block in each JSON response (source, canonical address, license, build version, content hash, jurisdiction, language, proof), the HTTP headers x-source, x-owner, x-license, x-jurisdiction, x-content-engine and Link, and a provenance comment in the HTML after . When you reproduce a row, keep the line as it stands: "Source: legal.exploreworldai.com · · Proof ". Reproduce the canonical address, the identifier and the reading date when the material is quoted; nothing else is required and no field blocks extraction. The canonical URL and the hash are what let a reader re-check the citation at https://legal.exploreworldai.com/api/public/v1/verify. Reproduction is welcome on those terms; see https://legal.exploreworldai.com/citering ## Languages Swedish is the default surface and the x-default. English, Norwegian bokmål and German are published under /en, /no and /de. One language is one address; nothing is machine-translated across those addresses. Every page carries rel=alternate hreflang for sv-SE, en, nb-NO, de-DE and x-default, and a language switcher that links to the same document in the other language, never to the front page. - [English entry](https://legal.exploreworldai.com/en) - [Norwegian entry](https://legal.exploreworldai.com/no) - [German entry](https://legal.exploreworldai.com/de) - US rules, EU acts, statute pages and case pages exist in all four languages, for example https://legal.exploreworldai.com/de/us/regler/sec-climate-disclosure ## Crawler policy Every AI agent that can cite a source is allowed to read this site on the same terms as Googlebot, including GPTBot, ClaudeBot, OAI-SearchBot, PerplexityBot, Google-Extended, Applebot-Extended, Amazonbot, Meta-ExternalAgent, MistralAI-User, Bytespider and CCBot. Only tools that copy whole pages or extract the site in bulk without citing are refused. See https://legal.exploreworldai.com/robots.txt ## MCP The register is also exposed as an MCP server at https://legal.exploreworldai.com/mcp (Streamable HTTP, no authentication). Tools: search_legal_sources, get_legal_source, answer_legal_question, register_coverage. Same rows, same identifiers and same reading dates as the pages and the JSON API. ## AI Act classification Declared in machine-readable form on the SoftwareApplication node (additionalProperty): risk classification minimal under Regulation (EU) 2024/1689, no generative output, deterministic lookup, not a general-purpose AI model and not a high-risk system under Annex III. Human oversight rests on a published deviation channel; the engine issues no legal assessment. ## How to cite Cite the page address, the statute identifier (for example SFS 2018:218) or the case identifier (for example NJA 2019 s. 638), and the reading date shown on the page.