How we work
How we make AI Act information verifiable
Responsible AI Platform translates official AI Act sources into practical interpretation, actions and evidence. The official source always prevails. Every publication shows its statement type, status, check date and review method and remains traceable through version history. The public knowledge layer is free and accessible without an account.
From official source to demonstrable action
Editorial principles
- Traceable to source
- A legal claim links to the official source and, where possible, the relevant article, paragraph, annex or passage.
- Verifiable process
- Review metadata shows which source and release checks were performed. A separate human legal review is only stated when it has been demonstrably recorded.
- Versioned
- We separately record what applied in law and what Responsible AI Platform had published at that time.
- Correctable
- A substantive error is not silently overwritten. Material corrections receive a visible note in the version history.
Official source, interpretation and recommended action are not the same
We label each statement type so a reader or AI agent can determine the authority it carries.
Official fact
Text, status, date or publication from EUR-Lex, the European Commission, the AI Office, a regulator or another named primary source.
RAIP interpretation
Our source-grounded explanation of meaning, scope or practical impact. It is not official guidance unless the cited source explicitly is.
Recommended action
A practical step or evidence measure we consider appropriate. A recommendation is not presented as a legal obligation.
If our summary and the official source differ, the official source prevails. Drafts, political agreements, guidance and binding law each receive a distinct status.
The publication check
An item receives review metadata only after the following source and release checks have demonstrably been completed:
- 1verify the identity, date, version and status of the primary source;
- 2verify the legal basis, affected roles, conditions and exceptions;
- 3label official fact, RAIP interpretation and recommended action separately;
- 4record the effective date and knowledge date;
- 5open and verify source locators and links;
- 6state material uncertainty or a source conflict visibly.
Source monitors, search technology and language models may support editorial work. Review metadata names the actual method used and is not an independent legal certification. Where no separate human final review is recorded, we do not claim one.
Two time axes prevent false certainty
effective_at- The date on which a rule, status or obligation applies in law for the requested snapshot.
known_at- The knowledge cut-off for information Responsible AI Platform had published and verified for that snapshot.
This allows an agent to ask not only what applies now, but what applied on an earlier date and which interpretation was then available. New information does not overwrite the historical snapshot.
Report a correction
Found an error, outdated status or missing primary source? Send us the URL, relevant passage, your explanation and preferably the primary source.
Report an editorial correction1. Record
We register the report and preserve its provenance.
2. Verify
We compare the claim with the primary source and applicable version.
3. Correct
If incorrect, we amend the current publication. A material change receives an explanation and version reference.
4. Respond
When contact details are provided, we explain how the report was handled.
Funding and interests
Responsible AI Platform is an independent initiative founded by Zahed Ashkara. Its public knowledge layer is currently operated from the founder’s own resources and has no paid subscription. Zahed also founded consultancy Embed AI and learning platform LearnWize. These related interests are disclosed here and provide no right to paid placement, a more favourable legal conclusion or influence over source selection.
If we later add sponsorship, data licensing or partnerships, we will identify the sender and relationship. Commercial support never changes the outcome of editorial review.
Free and publicly accessible
Reading, searching, checking sources and using the public data interface require no account. We do not require personal data as a condition for viewing legal information. A future account may only add convenience, such as saving, following or collaborating.
Free access does not automatically mean every underlying source document is free of rights. Official and external documents remain subject to their own source terms. Our dataset metadata describes available distributions without suggesting broader rights.
Not legal advice and not an official authority
Responsible AI Platform is not an institution of the European Union or a regulator. The information is general and cannot account for every fact, contract, sector rule or national implementation choice in a particular situation. Always check the official source before making a decision and seek contextual legal review where appropriate.
The same safeguards for AI agents
Human pages and machine output derive from the same versioned information model. The public API returns deterministic data, not a legal answer generated on demand. Objects have stable identifiers and carry source, review and temporal information.
Version history of this methodology
Version 1.0
8 August 2026
First public version: statement separation, transparent publication checks, dual time axes, correction path, interests disclosure and agent discovery documented.