AI Access Policy
This page explains how
licensefoundry.com signals what AI systems may and may not
do with the content of this Site. The signals themselves are published
in machine-readable form at
licensefoundry.com/robots.txt
following the
IAB Tech Lab
Content-Signals protocol. This page restates them in human-readable
form and explains the reasoning.
licensefoundry builds infrastructure for content owners to assert machine-verifiable rights about their digital assets. The default position of this Site — block AI training, allow AI search, allow traditional indexing — reflects the same model we sell to customers. We are running our own marketing site under the same kind of rights signals we ask data providers to use for their datasets.
1. The signals, in plain English
| Signal | Setting | What it means |
|---|---|---|
search |
yes | Traditional search engines (Google, Bing, Yahoo, and similar) may index this Site to provide hyperlink-based search results, including short text excerpts. |
ai-input |
yes | AI answer engines (Perplexity, ChatGPT Search, Claude with browsing, Gemini live grounding, and similar) may fetch content from this Site at query time to ground their responses, provided they attribute the source. |
ai-train |
no | Content from this Site may not be collected, ingested, scraped, or otherwise processed for the purpose of training, fine-tuning, or otherwise improving any AI model, foundation model, or machine-learning system, regardless of whether the training run is commercial, research, or non-commercial in nature. |
2. Why we set it this way
licensefoundry sells the position that rights to use content for AI
training should be expressed explicitly, in machine-readable form,
and respected by the systems that ingest content. Holding that
position credibly requires us to express the same kind of signal on
our own site. The robots.txt declaration above is the simplest
currently-deployable machine-readable expression of those rights;
the same content terms, expressed as a signed credential in the
licensefoundry format, are now public — see the
compliance scoring framework at /framework/v1.0
for the methodology we score credentials against, and
the partner page for the SDK that issues
credentials in this format.
Allowing search and AI search while denying training is not a contradiction. Search engines and answer engines fetch content at query time to surface it to a human asking a specific question — the human still arrives at our Site (or a clearly attributed reference to it) and the relationship is preserved. Training is a bulk-ingestion process in which our content is absorbed into a model weight tensor with no attribution, no traffic, and no ongoing relationship to the originating site. We do not consent to the latter and we deny it explicitly.
3. Legal grounding
The restriction on AI training expressed at
licensefoundry.com/robots.txt
is an
express reservation of rights under Article 4 of
Directive (EU) 2019/790 (the Digital Single Market
Directive) on copyright and related rights. Article 4 establishes
that the text-and-data-mining exception in EU copyright law does
not apply where the rights holder has expressly opted out, including
through machine-readable means appropriate to online content.
Our robots.txt is that machine-readable opt-out.
Restrictions also apply under all other applicable copyright, database-rights, and contract regimes — including but not limited to the terms set out at our Terms of Use.
4. Specific user-agent blocks
In addition to the Content-Signals declaration, we explicitly
Disallow: / the following known AI training crawlers
(this list is maintained by our edge provider and updated as new
crawlers are identified): Amazonbot, Applebot-Extended, Bytespider,
CCBot, ClaudeBot, Google-Extended, GPTBot, meta-externalagent, and
others. Live list is available at
/robots.txt.
5. If you operate an AI training pipeline and want access
Contact us at info@licensefoundry.com with a brief description of:
- What content you would like to use (specific URLs or page categories)
- What it would be used for (training a model? evaluation? research benchmarking?)
- Attribution and reporting commitments you can make
- Whether the resulting model is commercial, research, or open-weight
We will respond with terms — typically including attribution, reporting requirements, and any restrictions on downstream redistribution. If terms are agreed, we will issue a credential in the licensefoundry format that records the agreement and serves as a verifiable record both parties can reference. This is, by design, the same flow we offer our customers: rights are expressed as a signed credential, not as a hand-shake or a buried PDF.
The credential we issue is independently scorable against our
public framework. You can audit the compliance posture of any
agreement we propose by calling
GET /v1/credentials/{id}/compliance-score on
sandbox.licensefoundry.com (or production once
live) — the breakdown shows which categories of the framework the
credential satisfies, with the legal anchors cited per category.
Methodology:
/framework/v1.0.
6. Compliance and verification
Any party can verify the current state of our content signals at any time:
curl -s licensefoundry.com/robots.txt
We do not commit to any specific stability of the signals over time — we may update them as the underlying standards evolve, as new AI crawlers are identified, or as our position on specific use cases develops. The "Last updated" date at the top of this page reflects the most recent change.
7. Contact
Questions about this policy, requests for training access, or notices of non-compliance with the signals can be sent to info@licensefoundry.com.