WHAT THE ARTICLE IS ABOUT
Voluntary adoption of high-risk standards by non-high-risk AI providers
Article 95 creates a framework for voluntary codes of conduct through which providers of AI systems that are below the high-risk threshold can choose to apply some or all of the technical requirements designed for high-risk systems. The AI Office and member states facilitate and encourage these codes, and the Commission may approve them — giving them a level of official recognition.
WHAT IT SAYS
AI Office facilitates, Commission approves, providers may join voluntarily
- The AI Office and member states shall encourage and facilitate the drawing up of codes of conduct intended to foster the voluntary application of some or all of Chapter III Section 2 requirements to AI systems other than high-risk systems
- Codes of conduct may also address ethics principles such as sustainability, accessibility, and diversity
- The AI Office and member states shall facilitate participation of SMEs including startups, civil society, and other stakeholders in the development of codes
- Codes are drawn up by individual providers or organisations representing them — the Commission may approve codes that meet specified criteria
- Approved codes may be used as evidence of responsible AI practices — though they do not create legal presumptions of compliance with mandatory requirements
- The Commission may evaluate the effectiveness of approved codes and recommend amendments where necessary
- Participation in approved codes of conduct may be taken into account by market surveillance authorities when assessing an operator’s overall approach to compliance
WHO IS AFFECTED
Providers of non-high-risk AI systems who want to demonstrate responsible practices
- Providers of AI systems below the high-risk threshold who wish to signal commitment to higher standards
- Industry associations and organisations developing codes on behalf of their members
- The AI Office and member states facilitating code development
- The Commission which may approve codes and evaluate their effectiveness
WHAT IT MEANS FOR SMES
A trust signal and procurement differentiator — without mandatory compliance costs
- Participation in approved codes of conduct allows you to demonstrate responsible AI practices even if your system is not high-risk and you face no mandatory compliance obligations
- For B2B AI providers: public sector and large enterprise customers increasingly require evidence of responsible AI practices in procurement — an approved code of conduct provides that evidence
- The facilitation of SME participation in code development is explicit — engage with your industry association to shape codes that are practical for smaller businesses rather than having standards set by large incumbents
- Voluntary codes do not create legal liability — participating in a code and then failing to follow it does not automatically create an enforcement risk, but it may affect reputation
- Monitor which codes the Commission approves — approved codes carry more credibility with customers and procurement teams than informal commitments
Related Articles
- Article 96 — Commission guidelines (the guidelines that support code development and application)
- Article 56 — GPAI codes of practice (the mandatory GPAI compliance mechanism — distinct from these voluntary codes)
- Article 9 through 15 — High-risk requirements (the requirements that may be voluntarily adopted under this article)
- Article 62 — SME measures (SME participation in code development is facilitated under both articles)
