Resources · Trust center

Trust center

Who reviews Sustin's relationships, what each review status means, how Sustin uses AI, what each release contains, and how to reach us about corrections or security.

Reviewers

People who may record a review decision on a relationship. A Human-reviewed relationship names its reviewer and the date.

Seph Ziling Wang, Founder and Director, SUSTIN C.I.C.

Human review

Suggested: Proposed by AI or by a person, not yet reviewed. Suggested relationships are not published on this site. Release 2026.6: 434.
Human-reviewed: A named reviewer has checked the relationship and recorded a decision, date and note. Release 2026.6: 70.
Approved: A second qualified reviewer, independent of the first, has confirmed it. Release 2026.6: 0; none yet.

AI pre-reviews are recorded as such and never set a review status (DR-S015).

How we use AI

AI proposes; people decide. Every AI output is labelled with its method and tool in Sustin's data and exports. AI capabilities are described only as far as they exist, each marked Live, Beta, Coming soon or Planned.

AI-assisted mapping Live
AI proposes relationships between frameworks; named reviewers decide. In release 2026.6, 22 of the 70 published relationships began as AI-assisted proposals and 48 as expert proposals; each was reviewed by a named person.
Change monitoring Coming soon
Automated checks for new editions and amendments, reviewed before publication. A prototype reads official sources such as EUR-Lex and lists the editions, provisions and relationships a change affects; it never changes a reviewed record.
Request analysis Sample analysis
The Analyse tab on the home page shows a sample analysis computed from the release, without AI. In the AI version, which opens later, AI reads your request and structures its questions; matching, the value-chain cap and applicability come from the rules and reviewed mappings of the release, and every AI-written passage is tagged "AI-generated".
Sustin assistant Planned
Ask about your evidence, requests and frameworks.
Sustin MCP Planned
Connect AI assistants to Sustin's framework data.

AI Act transparency

Sustin's position on the AI Act (Regulation (EU) 2024/1689, as amended by Regulation (EU) 2026/1744):

AI literacy (Art. 4): SUSTIN C.I.C. supports the AI literacy of the staff and reviewers who work with Sustin.
Live AI interaction (Art. 50(1)): people are told when they interact with AI. The request analysis chat, the Sustin assistant, is labelled “AI” and opens with “You are chatting with an AI assistant.” and every AI-written passage is tagged “AI-generated”; the worked example is preset.
Machine-readable provenance (Art. 50(2)): AI-generated text produced by Sustin carries its method and tool in its data and exports.
Human review (Art. 50(4)): published AI-assisted content passes human review under the editorial responsibility of SUSTIN C.I.C.

This position is reviewed when codes of practice or implementing acts under Article 50(7) are adopted.

Releases

The registry is published in numbered releases, each with a manifest listing every file and its SHA-256 hash. A release is never changed after publication; this site is built from one pinned release, checked against its manifest hash.

Release: 2026.6
Date: 28 Oct 2026
Registry as of: 28 Oct 2026
Schema: 0.9
Manifest SHA-256: 86958babef45f76223403ccba170f8e45abc1379df902b7412f8674eebfa12b8

Change notes for release 2026.6

Release 2026.6: changes for consumers

Schema version 0.9. Copied into the release as CHANGES.md. Earlier releases are unchanged; i5Benchmark keeps its pin on 2026.1 until it changes it.

Applicability rules (new record type)

Guidance on when a framework or edition is likely to concern an organisation, stored as registry data (DR-S006) instead of site code. Each rule in data/applicability/ (released as data/applicability.json) has:

the framework (framework_id) and, where it rests on one, the edition (edition_id);
conditions: countries (ISO 3166-1 alpha-2 codes, or EU for every member state), size_bands (up-to-10, 11-to-1000, over-1000 employees) and value_chain_roles (supplier, reporting-company); an absent condition matches any value;
basis in own words, source_url and source_locator;
review_status: draft or reviewed. A reviewed rule names a reviewer listed in docs/reviewers.md, the date and a note; drafting by AI never sets reviewed (DR-S015).

Rules are guidance, not legal advice or a compliance assessment. Consumers show draft rules as draft guidance.

15 rules, all draft, for ESRS, the EU Taxonomy, VSME (two), CSDDD (two), Directive (EU) 2024/825, UK SRS, California SB 253, the SSBJ Standards, Australia's Chapter 2M and AASB S2, China's trial standards, IFRS S1 and S2, and the GRI Universal Standards. The thresholds of the CSRD, the value-chain cap and the CSDDD were read from Directive (EU) 2026/470 and the application date of Directive (EU) 2024/825 from its Article 4(1), both on the Publications Office (Cellar), 28 October 2026. Owner review: docs/review/applicability-rules.md.

Data

Frameworks, editions, provisions, concepts and mappings are unchanged from 2026.5: 102 frameworks, 122 editions, 70 reviewed and 434 suggested mappings. Applicability rules: 15 (new).

Exports Coming soon

Each release exports its relationships in SSSOM, an open exchange format, in two views: a consumer view with Human-reviewed and Approved relationships, and a research view with all statuses. Downloads from this site are in development.

i5benchmark-topic.sssom.tsv: licence at i5benchmark.com/licence/
sustin-core.sssom.tsv: licence at sustin.ai/terms

An open format does not change the licence of the data or of third-party content.

Content rights and licences

EU legal text is shown with attribution to EUR-Lex.
Licensed standards, such as ISO, IFRS and GRI, appear only as identifiers and Sustin's own summaries.
Sustin's own data and summaries: see the Terms of use (draft).
Fonts: Inter Tight and JetBrains Mono, SIL Open Font License 1.1. Map data: Natural Earth (public domain). Pixel illustrations: original, made for this site.

Legal pages: Privacy notice · Terms of use · Legal notice

Corrections and security reports

To report an error in the registry or on this site, or a security issue, write to [email protected]. For a correction, name the page, the framework or relationship and what is wrong; for a security report, describe the issue and how to reproduce it.