Frequently asked questions
Questions someone asks when discovering this tool for the first time.
What does Validator establish in relation to AI regulatory compliance?
Validator verifies document integrity and the technical conformance of a declaration with the PTC protocol. That evidence can support a transparency, audit or compliance process.
It does not by itself decide whether a specific case meets every obligation under the EU AI Act or another law. Article 50 applies to specific situations and actors and may also require visible notices, technical marking or other measures.
In Spain, Organic Bill 121/000096 remains before Parliament as of 21 July 2026. Validator should not be presented as a general legal certification, but as a verifiable tool for one specific part of the process: document traceability.
What is the difference between detection and declaration?
Detection tools try to identify whether a text was generated by AI without the author's knowledge or consent. This system starts from the opposite premise: someone who uses AI can choose to declare exactly how they did so, and that declaration can be cryptographically verified. Detection assumes adversarial intent. Declaration assumes voluntary accountability.
What stops someone from lying in a PTC?
Nothing technical prevents it. What the system guarantees is that the lie is permanently and immutably documented. If someone declares 'full human review' and the document later circulates without that review, the PTC preserves the original declaration and that discrepancy is traceable. The system does not prevent fraud: it makes fraud visible and durable.
What does the ethical context hash add?
The hash_ethical_context_sha256 hashes the declared creation conditions: who created the document, with which system, in which mode, for which purpose. If the document text stays unchanged but someone tries to retroactively claim different conditions (another system, another review level), the ethical hash will not match. The conditions of creation are as immutable as the content.
Can I modify the document after generating the PTC?
Yes, but validation will fail. The document hash is calculated over the text at the moment of generation. Any subsequent change, however small, produces a different hash. If you modify the document, you need to generate a new PTC.
What happens if I change the declared AI system but the text stays the same?
The document hash will match (the text is identical), but the ethical context hash will not. That is why two complementary hashes exist. The system detects that the declared creation conditions have changed even if the content is the same.
What does the document look like with the PTC included?
Validator preserves the source file separately and generates the canonical TXT artefact, consisting of the normalised textual object followed by the <PTC>…</PTC> block. When a reliable output adapter exists, it also generates a derived representation in the input format with the PTC incorporated.
How does someone else validate my declaration?
You can upload the canonical TXT/MD artefact or the distribution package. The system extracts the final PTC block, reconstructs the canonical object and recalculates all three hashes. Normative artefact validation does not require the original source; the package may preserve it to check reproducibility of the transformation.
What does local validation mean exactly?
All processing happens in your browser. The document is not sent to any server. Hashes are computed on your machine. There is no record of what you validate or generate. You can use the tool offline once the page has loaded.
Why TXT or MD rather than PDF or DOCX directly?
TXT and Markdown are Unicode text and allow reproducible validation without a document-extraction step. PDF and DOCX require adapters that may produce different results across tools or versions. Validator accepts them as experimental formats, preserves the source and always provides canonical TXT as the normative reference.
The PTC distribution package
What is the PTC distribution package and what does it contain?
What does the ZIP contain for each input format?
The common structure is the same; derived representations depend on the input format:
| You upload | Your distribution package contains |
|---|---|
| TXT | source/original.txt · canonical/document.txt · canonical/document_with_ptc.txt · ptc.txt · manifest.json · README.txt |
| MD | The above + representations/original_with_ptc.md |
| DOCX | DOCX source + two canonical TXT files + derived DOCX representation when self-validation succeeds + auxiliary files |
| PDF source + two canonical TXT files + auxiliary files. This version does not generate a PDF representation with PTC. | |
| CSV + URL | canonical/document.txt · canonical/document_with_ptc.txt · ptc.txt · manifest.json · README.txt (no documentary source) |
What can I upload to the validator and what result will I get?
| You upload to Validator | Result |
|---|---|
| ZIP | Validates the canonical artefact and checks the coherence of available pieces. |
| TXT | Direct canonical artefact validation. Reproducible result. |
| MD | Direct textual validation; Markdown syntax is part of the content. |
| DOCX | Experimental. Validator reconstructs the PTC from final paragraphs; for maximum reproducibility use the ZIP or canonical TXT. |
| Experimental. Extraction depends on the PDF adapter. For maximum reproducibility use the ZIP or canonical TXT. |