Article 95: codes of conduct for voluntary application of specific requirements
The AI Office and the Member States encourage and facilitate the drawing up of codes of conduct, including related governance mechanisms, intended to foster the voluntary application to AI systems other than high-risk AI systems of some or all of the requirements set out in Chapter III, Section 2, taking into account the available technical solutions and industry best practices. They also facilitate codes on the voluntary application of specific requirements to all AI systems, on the basis of clear objectives and key performance indicators, with elements such as the Union ethical guidelines for trustworthy AI, assessing and minimising the impact on environmental sustainability, promoting AI literacy, facilitating an inclusive and diverse design, and assessing and preventing the negative impact on vulnerable persons. Codes may be drawn up by individual providers or deployers or by organisations representing them, and in encouraging and facilitating them the specific interests and needs of SMEs, including start-ups, and of small mid-cap enterprises are taken into account; that last group was added by Article 1, point (35), of Regulation (EU) 2026/1744. This article imposes no duty on the organisation and replaces no obligation.
Paragraph 1 provides: the AI Office and the Member States shall encourage and facilitate the drawing up of codes of conduct, including related governance mechanisms, intended to foster the voluntary application to AI systems, other than high-risk AI systems, of some or all of the requirements set out in Chapter III, Section 2 taking into account the available technical solutions and industry best practices allowing for the application of such requirements.
Praxikon tracks Article 95: codes of conduct for voluntary application of specific requirements under the EU AI Act, checked against the official source on 6 September 2026, citing the source for every statement.
- Status
- Applicable
- Application date
- Not recorded
- Version
- 1.0.0
- Last reviewed
- 6 September 2026
Review status: placed against the official source (6 September 2026). Next check due by 5 March 2027. The check date is the knowledge date of this version; no later recheck has been recorded.
From source to evidence
Why this obligation applies, what it asks of you, and what you show for it.
Applies
Applicable · Not recorded
For whom
- Deployer
- Provider of an AI system
What you do
Scope a voluntary code of conduct and separate it from your duties
What you record
Register of voluntary commitments alongside the obligations
Official source
Who this is relevant to
When this applies
Deployer
An organisation using an AI system under its authority, excluding personal non-professional use.
Provider of an AI system
A party that develops or has an AI system developed and places it on the market under its own name.
- 1Paragraph 1 expressly concerns AI systems other than high-risk AI systems, and the voluntary application to them of some or all of the requirements set out in Chapter III, Section 2. Paragraph 2 is wider and concerns the voluntary application, including by deployers, of specific requirements to all AI systems, on the basis of clear objectives and key performance indicators to measure the achievement of those objectives. Article 95 sits in Chapter X, which is named in none of the three exceptions in the third paragraph of Article 113; the general date of application in the second paragraph therefore governs, 2 August 2026.
- 2Paragraph 3 sets out who may draw up a code of conduct: individual providers or deployers of AI systems, organisations representing them, or both, including with the involvement of any interested stakeholders and their representative organisations, including civil society organisations and academia. A code may cover one or more AI systems, taking into account the similarity of the intended purpose of the relevant systems.
What the official source establishes
Paragraph 3 provides: codes of conduct may be drawn up by individual providers or deployers of AI systems or by organisations representing them or by both, including with the involvement of any interested stakeholders and their representative organisations, including civil society organisations and academia. Codes of conduct may cover one or more AI systems taking into account the similarity of the intended purpose of the relevant systems. Paragraph 4, as replaced by Article 1, point (35), of Regulation (EU) 2026/1744, provides: the AI Office and the Member States shall take into account the specific interests and needs of SMEs, including start-ups, and small mid-cap enterprises, when encouraging and facilitating the drawing up of codes of conduct.
Our interpretation
The official source remains authoritative. This general interpretation is not legal advice.
Read this article first for what it is not. It imposes no duty on your organisation: paragraphs 1, 2 and 4 address the AI Office and the Member States, and paragraph 3 only says who may draw a code up. And paragraph 1 is expressly about AI systems other than high-risk AI systems. That is the core of the provision and at the same time the source of the biggest misunderstanding in practice. The common thought is that a code of conduct can replace a mandatory requirement, or that signing one produces a form of compliance a regulator can hold against you or count in your favour. Neither is in the text. The voluntariness in Article 95 concerns requirements that precisely do not apply to your system: Chapter III, Section 2, contains the requirements for high-risk AI systems, and Article 95 invites you to apply some or all of those requirements to a system that is not high-risk. The reverse, a lighter reading of a requirement that does apply, is not on offer in this article. Two limits go with that. Article 4 requires providers and deployers to take measures to ensure a sufficient level of AI literacy. Article 95(2), point (c), names promoting AI literacy as a possible element of a code of conduct. That is not a duplication: the first is the duty, the second is the superstructure. A code of conduct on AI literacy is therefore a fine addition and never a replacement of Article 4. The same holds for transparency: Article 50 imposes a number of transparency duties, and a code that says something about them leaves Article 50 untouched. Anyone suggesting in a tender document or annual report that a code covers one of those duties is making a claim the text does not support.
Then read the article for what it does offer, because that is underestimated. This is the only place in the Regulation that says what an organisation can do of its own accord, and the enumeration in paragraph 2 is strikingly concrete. It names assessing and minimising the impact of AI systems on environmental sustainability, including as regards energy-efficient programming and techniques for the efficient design, training and use of AI. That topic appears nowhere else in the Regulation as something you can do, and for many organisations it is the topic a board is already asking about. It further names facilitating an inclusive and diverse design, including through inclusive and diverse development teams and the involvement of stakeholders in that process, and assessing and preventing the negative impact on vulnerable persons or groups, including as regards accessibility for persons with a disability and gender equality. Note the form paragraph 2 prescribes alongside, because it separates a serious code from a statement of intent: clear objectives and key performance indicators to measure the achievement of those objectives. A code without measurable indicators does not meet the form the provision itself describes. In practice this is where you can hook your own AI policy onto the Regulation without promising anything you cannot deliver. The most useful use is the inverse application of paragraph 1: for a system that is not high-risk, deliberately adopt part of Chapter III, Section 2, for example the keeping of logs or the documentation of data quality, and say which part you are not adopting and why. That is defensible and it is preparation as well, because a system sometimes changes classification.
What you can do now
First make the separation visible in your own documentation: set out in two columns what you must do and what you do voluntarily, and make sure no line from the second column is used as cover for a line from the first. Then test every existing ethical code, AI policy document or supplier promise against that separation, because that is usually where the blending sits. If you are considering a code under Article 95, choose deliberately between the two routes: the paragraph 1 route for a system that is not high-risk, where you name which requirements of Chapter III, Section 2, you adopt and which you do not, or the paragraph 2 route for specific requirements across all your AI systems. In both cases give the code the form paragraph 2 describes: clear objectives and key performance indicators with which you measure the achievement, with a named owner and a moment of measurement. Then pick the elements that genuinely mean something to you rather than all five; environmental sustainability, AI literacy, inclusive and diverse design and the impact on vulnerable persons are separate topics with separate data. If you are an SME or a start-up, ask your national competent authority or the AI Office what support for drawing one up is available, because paragraph 4 requires your interests and needs to be taken into account. Finally, put no wording in quotations, tender responses or annual reports from which a reader could infer that participation in a code of conduct covers an obligation under the Regulation.
- 01
Scope a voluntary code of conduct and separate it from your duties
Choose the paragraph 1 or the paragraph 2 route, name which requirements you apply voluntarily and which you do not, give the code clear objectives and key performance indicators, and keep the voluntary commitments administratively separate from the obligations that continue to apply in full.
What to retain
Register of voluntary commitments alongside the obligations
Per system and per commitment: which requirements you apply voluntarily, under which paragraph of Article 95, with which objectives and key performance indicators, who owns it, when it is measured, and which obligation under the Regulation continues to apply in full alongside it.
Control and reassessment
Review that keeps voluntary and mandatory apart
The control that no external statement, quotation, tender response or annual report presents a code of conduct as cover for an obligation under the Regulation, and that every voluntary commitment has an owner, an indicator and a moment of measurement before it goes out.
Public tools
Full text of Article 95
The full legal text in the public AI Act Explorer.
Conditions and exceptions
- Article 95 contains no provision that sets aside, suspends or replaces an obligation under this Regulation. The word voluntary refers to the application of requirements that precisely do not apply to the system concerned, and not to the requirements that do apply. A code of conduct on AI literacy leaves Article 4 untouched, and a code of conduct on transparency leaves Article 50 untouched.
Official sources and locators
EU Artificial Intelligence Act 2024/1689
European Parliament and Council | original-oj-2024-07-12
Source locator: Article 95(1) to (4)
EU Artificial Intelligence Act 2024/1689
European Parliament and Council | original-oj-2024-07-12
Source locator: Article 4; Article 50; Chapter III, Section 2; Article 113, second paragraph
Referring to this object
Citation block
Copy this reference into your advice, article or file. The identifier, the version and the hash keep the statement findable later, even once the dataset has moved on.
Reference
Praxikon, "Article 95: codes of conduct for voluntary application of specific requirements", praxikon:eu:ai-act:obligation:article-95-voluntary-codes-of-conduct@1.0.0, dataset praxikon:sys:registry:dataset:ai-act-implementation-graph 2.2.0 (schema 1.5.0), effective_at 2026-08-02T00:00:00.000Z, known_at 2026-09-06T00:00:00.000Z, sha256 2b71546d72d3dbc5918daae97128ae05900c52ce93ba71de9866e90de4176e14, https://www.praxikon.com/en/verplichtingen/article-95-voluntary-codes-of-conduct (https://www.praxikon.com/api/v1/obligations?id=praxikon%3Aeu%3Aai-act%3Aobligation%3Aarticle-95-voluntary-codes-of-conduct&effective_at=2026-08-02&known_at=2026-09-06&lang=en, accessed 2026-09-15)
Short form
praxikon:eu:ai-act:obligation:article-95-voluntary-codes-of-conduct@1.0.0 (sha256 2b71546d)
BibTeX
@misc{praxikon-eu-ai-act-obligation-article-95-voluntary-codes-of-conduct-1-0-0,
author = {{Praxikon}},
title = {Article 95: codes of conduct for voluntary application of specific requirements},
year = {2026},
version = {1.0.0},
number = {praxikon:eu:ai-act:obligation:article-95-voluntary-codes-of-conduct},
howpublished = {AI Act Change \& Evidence Graph, dataset 2.2.0, schema 1.5.0},
note = {effective_at 2026-08-02T00:00:00.000Z; known_at 2026-09-06T00:00:00.000Z; sha256 2b71546d72d3dbc5918daae97128ae05900c52ce93ba71de9866e90de4176e14},
url = {https://www.praxikon.com/en/verplichtingen/article-95-voluntary-codes-of-conduct},
urldate = {2026-09-15},
language = {en}
}CSL JSON
[
{
"id": "praxikon:eu:ai-act:obligation:article-95-voluntary-codes-of-conduct@1.0.0",
"type": "dataset",
"title": "Article 95: codes of conduct for voluntary application of specific requirements",
"container-title": "AI Act Change & Evidence Graph",
"publisher": "Praxikon",
"version": "1.0.0",
"number": "praxikon:eu:ai-act:obligation:article-95-voluntary-codes-of-conduct",
"URL": "https://www.praxikon.com/en/verplichtingen/article-95-voluntary-codes-of-conduct",
"language": "en",
"issued": {
"date-parts": [
[
2026,
9,
6
]
]
},
"accessed": {
"date-parts": [
[
2026,
9,
15
]
]
},
"note": "dataset praxikon:sys:registry:dataset:ai-act-implementation-graph 2.2.0; schema 1.5.0; effective_at 2026-08-02T00:00:00.000Z; known_at 2026-09-06T00:00:00.000Z; sha256 2b71546d72d3dbc5918daae97128ae05900c52ce93ba71de9866e90de4176e14; retrieved_from https://www.praxikon.com/api/v1/obligations?id=praxikon%3Aeu%3Aai-act%3Aobligation%3Aarticle-95-voluntary-codes-of-conduct&effective_at=2026-08-02&known_at=2026-09-06&lang=en; licence https://www.praxikon.com/nl/legal/terms"
}
]How to verify a reference later is set out in the methodology. Terms
Version history
v1.0.0
2 August 2026
Article 95: codes of conduct for voluntary application of specific requirements
The AI Office and the Member States encourage and facilitate the drawing up of codes of conduct, including related governance mechanisms, intended to foster the voluntary application to AI systems other than high-risk AI systems of some or all of the requirements set out in Chapter III, Section 2, taking into account the available technical solutions and industry best practices. They also facilitate codes on the voluntary application of specific requirements to all AI systems, on the basis of clear objectives and key performance indicators, with elements such as the Union ethical guidelines for trustworthy AI, assessing and minimising the impact on environmental sustainability, promoting AI literacy, facilitating an inclusive and diverse design, and assessing and preventing the negative impact on vulnerable persons. Codes may be drawn up by individual providers or deployers or by organisations representing them, and in encouraging and facilitating them the specific interests and needs of SMEs, including start-ups, and of small mid-cap enterprises are taken into account; that last group was added by Article 1, point (35), of Regulation (EU) 2026/1744. This article imposes no duty on the organisation and replaces no obligation.
Corrections to this obligation
No substantive correction to this object has been recorded.
Open the correction log
Help with implementation
Zahed Ashkara, jurist and freelance AI & Privacy Consultant, supports implementation with your team through Embed AI.
View AI governance at Embed AIFor AI agents and integrations
This page and the machine output derive from the same versioned object. Use the API for deterministic filters by role, topic and time.