Choose an institution to see its records and dossier, or choose a topic.
Through year 2027 · all records, including scheduled dates 2026 2025 2024 2023 2022 2021 2020 2019 2018 2017 2016 2015 2014 2013 2012 2011 2010 2009 2008 2007 2006 2005 2004 2003 2002 2001 2000 1999 1998 1997 1996 1994 1993 1992 1991 1989 1986 1985 1984 1983 1979 1966 1965 1961 1960 1955 1952 1950 1872 1863 Show Organisations and infrastructure All types Organisation Person Document Concept Infrastructure Public body Country Reset filters
101 people, organisations and other records · 133 relationships · 34 claims about a valueEverything recorded up to the chosen year, whether or not still in effect · Undated records included
Claim and evidence Close ×
The final Code's Appendix 1.4 specifies four systemic risks (CBRN, loss of control, cyber offence, harmful manipulation), where the initial draft's Measure 6.1 listed six types, including automated use of models for AI research and development and large-scale discrimination, which are not among the final four. Copy claim link
Copy claim reference
Machine-checked, not reviewed by an editor Inferred Supported
Supporting evidence: primary · 2 sources
Reasoning and premises Both lists are quoted at their locators; the comparison is a set difference of the listed headings.
Basis of the change Inferred by comparing the two editions. No publisher statement of this change is recorded. Subject version Final version (10 July 2025) · Published 2025-07-10 · Effective date not established Object version Initial draft (14 November 2024) · Published 2024-11-14 · Effective date not established Date 2025-07-10 · publication Scope and limits Compares the headings of the two lists. The two types are not deleted from the final Code: its description of loss of control includes autonomously creating or improving AI models (Appendix 1.4(2)); Appendix 1.3 lists capabilities to automate AI research and development as a source of systemic risk; Appendix 1.1 lists risks to non-discrimination among the types of risk to draw on. The record says only that the two types are not among the four specified risks. The intermediate drafts were not retrieved.
The evidence 04 primary source Supports the claim
General-Purpose AI Code of Practice: Safety and Security chapter (final version, 10 July 2025) ↗ European Commission (AI Office) · 2025-07-10
Archived copy, captured 2025-07-10: General-Purpose AI Code of Practice: Safety and Security chapter (final version, 10 July 2025) ↗
Copy source reference
Cite the original source Cite the original document, not Selekoa. Check these details against it.
Copy APA
Download BibTeX Download RIS Download CSL-JSON
Source excerpt “(2) Loss of control” Location in source Appendix 1.4, points (1) to (4) (p. 37)
How it was read Retrieved directly from the publisher
Machine-checked · 2026-10-28
Source details and check notes Authors Chairs and vice-chairs of the Code’s working groups (independent experts) Version PDF served by the Commission on 2026-10-04 (a re-rendered file of 43 pages); it has the same 15,392 words as the Internet Archive capture of 2025-07-10 (10:58 UTC, 40 pages), with a different reading order on the cover page. Every quote matched in both expressions. Canonical address Open the canonical source ↗ Accessed 2026-10-04 Language English (en) Archive copy Open archived copy, captured 2025-07-10: General-Purpose AI Code of Practice: Safety and Security chapter (final version, 10 July 2025) ↗ . Captured on the recorded publication date. An archive copy shows only what the archive retrieved on 2025-07-10. It does not prove what the page said earlier, and it can differ from the copy checked on 2026-10-04.Attribution General-Purpose AI Code of Practice, Safety and Security chapter (10 July 2025), published by the European Commission. Reuse European Union; quoted briefly under the editorial policy. Selekoa practice: metadata, attributed paraphrases and short attributed quotations under the editorial policy (at most 50 words each; at most 400 quoted words per document across the dataset, and about 10% of a document under about 2,000 words). Documents are linked externally, not redistributed. Publisher terms that forbid quotation override this practice. Check notes Candidate pack passage, frameworks to law (research/frameworks-to-law). The quotation (4 words) matched the text retrieved for the candidate pack on 2026-10-27 and again at admission on 2026-10-28, in a fresh retrieval whose SHA-256 equals the pack’s. Read directly from ec.europa.eu. Matching ignored differences of whitespace, case, quotation marks, dashes and Unicode compatibility forms, so the source’s typography can differ. Headings of points (1)-(4): Chemical, biological, radiological and nuclear; Loss of control; Cyber offence; Harmful manipulation. A cross-family machine check (2026-10-28, item 25) answered the record’s question blind from fresh retrievals and agreed with the record. No human editorial review. primary source Supports the claim
General-Purpose AI Code of Practice: Safety and Security chapter (final version, 10 July 2025) ↗ European Commission (AI Office) · 2025-07-10
Archived copy, captured 2025-07-10: General-Purpose AI Code of Practice: Safety and Security chapter (final version, 10 July 2025) ↗
Copy source reference
Cite the original source Cite the original document, not Selekoa. Check these details against it.
Copy APA
Download BibTeX Download RIS Download CSL-JSON
Source excerpt “(4) Harmful manipulation” Location in source Appendix 1.4, point (4) (p. 37)
How it was read Retrieved directly from the publisher
Machine-checked · 2026-10-28
Source details and check notes Authors Chairs and vice-chairs of the Code’s working groups (independent experts) Version PDF served by the Commission on 2026-10-04 (a re-rendered file of 43 pages); it has the same 15,392 words as the Internet Archive capture of 2025-07-10 (10:58 UTC, 40 pages), with a different reading order on the cover page. Every quote matched in both expressions. Canonical address Open the canonical source ↗ Accessed 2026-10-04 Language English (en) Archive copy Open archived copy, captured 2025-07-10: General-Purpose AI Code of Practice: Safety and Security chapter (final version, 10 July 2025) ↗ . Captured on the recorded publication date. An archive copy shows only what the archive retrieved on 2025-07-10. It does not prove what the page said earlier, and it can differ from the copy checked on 2026-10-04.Attribution General-Purpose AI Code of Practice, Safety and Security chapter (10 July 2025), published by the European Commission. Reuse European Union; quoted briefly under the editorial policy. Selekoa practice: metadata, attributed paraphrases and short attributed quotations under the editorial policy (at most 50 words each; at most 400 quoted words per document across the dataset, and about 10% of a document under about 2,000 words). Documents are linked externally, not redistributed. Publisher terms that forbid quotation override this practice. Check notes Candidate pack passage, frameworks to law (research/frameworks-to-law). The quotation (3 words) matched the text retrieved for the candidate pack on 2026-10-27 and again at admission on 2026-10-28, in a fresh retrieval whose SHA-256 equals the pack’s. Read directly from ec.europa.eu. Matching ignored differences of whitespace, case, quotation marks, dashes and Unicode compatibility forms, so the source’s typography can differ. A cross-family machine check (2026-10-28, item 25) answered the record’s question blind from fresh retrievals and agreed with the record. No human editorial review. primary source Supports the claim
First Draft General-Purpose AI Code of Practice ↗ European Commission · 2024-11-14
Copy source reference
Cite the original source Cite the original document, not Selekoa. Check these details against it.
Copy APA
Download BibTeX Download RIS Download CSL-JSON
Source excerpt “Automated use of models for AI Research and Development” Location in source Measure 6.1, list (p. 17)
How it was read Retrieved directly from the publisher
Machine-checked · 2026-10-28
Source details and check notes Authors Chairs and Vice-Chairs of the Working Groups of the General-Purpose AI Code of Practice Version Retrieved on 2026-10-28 (SHA-256 e7522e16b0dd05f368592ff19cd2e4af1f1208be88015bfa0059cdb5f6a32354). The PDF served by the Commission's newsroom (document 109946), 36 pages, every page marked “DRAFT DOCUMENT”; its metadata gives creation on 14 November 2024, 10:55 UTC. Decision 5 of 2026-10-28: the Commission's own file is cited. The candidate pack first read the copy held by the Dutch Senate (Eerste Kamer der Staten-Generaal, https://www.eerstekamer.nl/9370000/1/j4nvi0xeni9vr2l_j9vvkyblr2b3ny8/vmidr7qz6qv4/f=/vmidr9wfudtt.pdf, SHA-256 605816c07298146ec981fb46929518ee09d3072ea49994d2b1b032366555ba87); the cross-family check of 2026-10-28 found the two files' extracted text identical on all 36 pages, although the raw files differ. Canonical address Open the canonical source ↗ Accessed 2026-10-28 Language English (en) Attribution Chairs and Vice-Chairs of the Working Groups of the General-Purpose AI Code of Practice, “First Draft General-Purpose AI Code of Practice”, 2024-11-14. Reuse No open redistribution licence asserted. Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution. Check notes Candidate pack passage, frameworks to law (research/frameworks-to-law). The quotation (9 words) matched the text retrieved for the candidate pack on 2026-10-27 and again at admission on 2026-10-28, in a fresh retrieval of the Commission’s own file, which this record cites (SHA-256 e7522e16b0dd05f368592ff19cd2e4af1f1208be88015bfa0059cdb5f6a32354). Read directly from the Commission's newsroom. Matching ignored differences of whitespace, case, quotation marks, dashes and Unicode compatibility forms, so the source’s typography can differ. A cross-family machine check (2026-10-28, item 25) answered the record’s question blind from fresh retrievals and agreed with the record. No human editorial review. primary source Supports the claim
First Draft General-Purpose AI Code of Practice ↗ European Commission · 2024-11-14
Copy source reference
Cite the original source Cite the original document, not Selekoa. Check these details against it.
Copy APA
Download BibTeX Download RIS Download CSL-JSON
Source excerpt “Large-scale discrimination: Large-scale illegal discrimination of individuals, communities, or societies.” Location in source Measure 6.1, list (p. 17)
How it was read Retrieved directly from the publisher
Machine-checked · 2026-10-28
Source details and check notes Authors Chairs and Vice-Chairs of the Working Groups of the General-Purpose AI Code of Practice Version Retrieved on 2026-10-28 (SHA-256 e7522e16b0dd05f368592ff19cd2e4af1f1208be88015bfa0059cdb5f6a32354). The PDF served by the Commission's newsroom (document 109946), 36 pages, every page marked “DRAFT DOCUMENT”; its metadata gives creation on 14 November 2024, 10:55 UTC. Decision 5 of 2026-10-28: the Commission's own file is cited. The candidate pack first read the copy held by the Dutch Senate (Eerste Kamer der Staten-Generaal, https://www.eerstekamer.nl/9370000/1/j4nvi0xeni9vr2l_j9vvkyblr2b3ny8/vmidr7qz6qv4/f=/vmidr9wfudtt.pdf, SHA-256 605816c07298146ec981fb46929518ee09d3072ea49994d2b1b032366555ba87); the cross-family check of 2026-10-28 found the two files' extracted text identical on all 36 pages, although the raw files differ. Canonical address Open the canonical source ↗ Accessed 2026-10-28 Language English (en) Attribution Chairs and Vice-Chairs of the Working Groups of the General-Purpose AI Code of Practice, “First Draft General-Purpose AI Code of Practice”, 2024-11-14. Reuse No open redistribution licence asserted. Bibliographic metadata, editorial paraphrases and short attributed quotations (at most 50 words each). Recheck terms before wider redistribution. Check notes Candidate pack passage, frameworks to law (research/frameworks-to-law). The quotation (10 words) matched the text retrieved for the candidate pack on 2026-10-27 and again at admission on 2026-10-28, in a fresh retrieval of the Commission’s own file, which this record cites (SHA-256 e7522e16b0dd05f368592ff19cd2e4af1f1208be88015bfa0059cdb5f6a32354). Read directly from the Commission's newsroom. Matching ignored differences of whitespace, case, quotation marks, dashes and Unicode compatibility forms, so the source’s typography can differ. A cross-family machine check (2026-10-28, item 25) answered the record’s question blind from fresh retrievals and agreed with the record. No human editorial review. Status and meaning Editorial status. Not reviewed by an editor . An exploratory record that no editor has accepted. This status does not say whether a person has read any linked passage. Each evidence record above shows how it was checked.
Relation. One scoped difference between document versions. The claim identifies publisher-stated change, official changelog/redline, or an explicitly inferred comparison. It does not establish implementation, effectiveness or a safety ranking.
Date. Dated by the final Code (10 July 2025); compared with the initial draft (14 November 2024).
Record ID fl-eu-cop-final-changes-from-draft1-risk-types