Our posture on copyright
NOSIBLE uses public pages to build search, not to republish articles. The copyright question is whether the product helps users find and evaluate sources or replaces the original work. NOSIBLE is built for discovery, attribution, and non-substitutive retrieval (Authors Guild v. Google).
- NOSIBLE builds search, rankings, metadata, timestamps, and snippets from public pages.
- Users get short snippets, source metadata, dates, and links back to originals.
- NOSIBLE is for discovery and evidence review, not article consumption.
- It does not sell cached pages, article archives, full text, or publisher replacement.
- Rate limits and abuse monitoring reduce full-work reconstruction risk.
- NOSIBLE respects robots.txt, rights reservations, and source-removal requests.
- Search cases support indexing for discovery: Authors Guild, HathiTrust, Kelly, and Perfect 10.
- Facts are not copyrightable. Hot-news claims remain narrow under Feist and NBA v. Motorola.
NOSIBLE uses public pages to build search, not to republish articles
Copyright law distinguishes technical copying for discovery from redistribution for consumption. Search and snippet cases protect transformative, limited, non-substitutive uses, including Authors Guild v. Google, HathiTrust, Kelly v. Arriba Soft, Perfect 10 v. Amazon, and A.V. v. iParadigms. NOSIBLE uses public pages for discovery, attribution, ranking, and source retrieval.
Google v. Oracle confirms that commercial use does not defeat fair use where the use adds a different function and does not substitute for the original market. Warhol v. Goldsmith reinforces the need to keep purpose different from the original work. NOSIBLE's purpose is search and source discovery, not article consumption.
Copyright risk is controlled by transformation, limits, and attribution
| Area | NOSIBLE position |
|---|---|
| Transformative use | Public pages become indexes, entity signals, timestamps, source metadata, shards, relevance scores, and ranked snippets. |
| Snippet limits | Returned text is limited to what is needed to judge relevance. It should answer whether the source matters, not replace the work. |
| Attribution | Every result identifies the source and routes the user back to the original page. |
| Non-substitution | NOSIBLE is designed for source discovery, not article consumption, archive reading, or full-text resale. |
| Opt-out handling | NOSIBLE respects robots.txt and machine-readable reservations, which supports implied-license and lawful-access arguments (Field v. Google; EU Directive 2019/790). |
| Anti-extraction | Retrieval endpoints are rate-limited and monitored so customers cannot reconstruct complete works at scale. |
| Facts and events | Facts are not copyrightable, and hot-news claims remain narrow (Feist; NBA v. Motorola). |
| Legal review | Enterprise terms and indemnity can be discussed based on the use case, deployment model, contract, and risk allocation. |
Snippets are better for AI systems and safer for publishers
Short snippets are a better technical unit for search, AI, and quantitative analysis. Full text often includes navigation, boilerplate, duplicate paragraphs, unrelated context, and long passages that dilute the signal. A snippet gives the model or analyst the relevant evidence and leaves the original source in control of the full work.
Search indexing, snippet view, temporary technical copying, very short extracts, hyperlinking, and structured fairness analysis are supported across major jurisdictions, including the U.S. search cases above, EU Directive 2019/790, EU Directive 2001/29/EC, UK CDPA section 28A, UK CDPA section 29A, and CCH Canadian Ltd. v. Law Society of Upper Canada. UK commercial text-and-data-mining law is narrower than the EU framework, so NOSIBLE does not rely on a blanket UK commercial TDM claim.
Common copyright questions
Does NOSIBLE republish articles?
No. NOSIBLE does not sell a reading product, archive product, cached-page product, or full-text resale product. It returns relevance-bound snippets, metadata, dates, and links back to the original source. Search cases protect tools that help users find and evaluate works without giving them a substitute for the works (Authors Guild v. Google).
What is NOSIBLE's copyright basis?
NOSIBLE is non-substitutive search. It transforms public pages into indexes, rankings, metadata, shards, relevance scores, and short previews so users can discover sources rather than consume original works inside NOSIBLE. HathiTrust treated full-text search as different from reading the underlying work.
Why are snippets important?
Snippets show enough text to judge relevance without replacing the original work. NOSIBLE returns short snippets of 256 words or less, with source attribution. That limit matters legally because it avoids a full-text reading experience, and technically because snippets give models and analysts relevant evidence without forcing them through boilerplate, navigation, duplicate text, and unrelated context.
Why does NOSIBLE say snippets are technically better than full text?
Snippets improve signal-to-noise, normalize document length, reduce semantic dilution in vector search, fit more sources into an AI context window, reduce latency and inference cost, and reduce hallucination risk by grounding generation in specific retrieved passages. The same design that reduces substitution risk also improves retrieval quality.
What prevents customers from reconstructing full works?
NOSIBLE uses anti-extraction controls, including strict rate limits and abuse monitoring on retrieval endpoints. The product is designed around snippets, source metadata, rankings, and links, not full-text reconstruction. Fair-use analysis turns heavily on non-substitution and market effect (Authors Guild v. Google; Google v. Oracle).
How do anti-extraction controls protect the search product?
Anti-extraction protects the search design. If customers could reconstruct full works through repeated retrieval, NOSIBLE would start to look like a redistribution system instead of a search system. Rate limits and abuse monitoring help keep retrieval focused on relevance, evidence, attribution, and source discovery instead of bulk text extraction.
How does attribution reduce copyright risk?
Every result includes the source domain, original URL, and date visited. Attribution matters because it routes human and AI traffic back to the original publisher for consumption. It also reinforces the discovery function of the product. NOSIBLE is not trying to keep users away from source websites. It is trying to help them find the right source faster.
Does NOSIBLE respect robots.txt and opt-out signals?
Yes. NOSIBLE respects robots.txt and machine-readable reservations, and source owners can request removal. Robots and no-index style controls support implied-license and opt-out analysis in search cases such as Field v. Google. EU text-and-data-mining rules also recognize machine-readable rights reservations under Directive 2019/790.
Is NOSIBLE a generative AI product that reproduces publisher text?
No. NOSIBLE is search and retrieval infrastructure. It returns dated, attributed source results and snippets so users can inspect evidence and open the original source. It does not generate substitute articles or expressive outputs from publisher text. That separates NOSIBLE from disputes centered on competing expressive outputs, including Thomson Reuters v. Ross and the pending OpenAI litigation.
Why does source attribution matter for AI agents?
Source attribution keeps the evidence visible. An AI agent can inspect the source domain, original URL, date visited, and relevant snippet before using a result. That reduces black-box retrieval risk and routes the user back to the original publisher when the full work needs to be read.
How is NOSIBLE different from a web archive or scraping vendor?
A web archive or bulk scraping product can create substitution risk if it gives users cached pages, full-text redistribution, or archive-style access to third-party content. NOSIBLE turns source pages into search infrastructure and returns limited snippets with attribution. Users find third-party articles in NOSIBLE and read them at the original source.
How is NOSIBLE different from Common Crawl?
NOSIBLE is a search and retrieval product, not a bulk web corpus for redistribution. It returns snippets, rankings, metadata, dates, source attribution, and links back to original pages. Common Crawl-style access can be useful for some technical teams, but NOSIBLE is built around discovery, point-in-time retrieval, and anti-extraction controls.
Does NOSIBLE provide cached full-page access?
No. NOSIBLE does not provide cached full-page access, archive-style reading, or full-text resale. The product returns snippets, metadata, rankings, dates, and links back to the source. Users discover the source in NOSIBLE and consume the full work at the original website.
How is NOSIBLE safer than naive scraping?
NOSIBLE is more conservative than products that provide cached pages, archive-style access, full-text redistribution, or bulk extraction of third-party text. It returns limited snippets with source attribution and retrieval controls. The product is built for discovery, not consumption of publisher content inside NOSIBLE.
Which U.S. cases support the copyright position?
Authors Guild v. Google, HathiTrust, Kelly v. Arriba Soft, Perfect 10 v. Amazon, A.V. v. iParadigms, and Field v. Google support the position. They protect search, indexing, snippets, thumbnails, anti-plagiarism indexing, caching, or source-location functions when the tool helps users find information without replacing the work.
What does the EU and UK framework add?
EU and UK frameworks cover technical copies, text and data mining, very short extracts, hyperlinking, database-right limits around substantial extraction, and the UK's temporary-copying exception in CDPA section 28A. The EU also has a broader commercial text-and-data-mining exception with opt-outs under Directive 2019/790. UK section 29A is narrower and non-commercial.
What about Canada, Japan, Singapore, Israel, and Australia?
The same pattern appears outside the U.S., EU, and UK. Canada treats fair dealing as a structured balancing test. Japan allows non-enjoyment uses including data analysis and machine learning. Singapore allows computational data analysis of lawfully accessed works if originals are not distributed. Israel and Australia also focus on transformation, proportionality, and market effect.
Can enterprise customers discuss indemnity?
Yes. Third-party intellectual property rights indemnity can be negotiated for enterprise customers. The terms depend on the use case, deployment model, contract, and risk allocation. Indemnity allocates commercial and legal risk between NOSIBLE and the customer.