Glossary · EU AI Regulation AI Pact (EU AI Pact (voluntary AI Act compliance framework))
Voluntary EU framework launched by the AI Office in 2024 enabling AI providers and deployers to commit to early AI Act compliance ahead of mandatory deadlines. Over 100 signatories as of 2026, including major foundation-model providers and European AI deployers.
## What the AI Pact actually is
The AI Pact is a voluntary EU initiative launched by the [AI Office](/en/glossary/ai-office/) in 2024 enabling AI providers, deployers, and other stakeholders to commit to early AI Act compliance ahead of mandatory regulatory deadlines. It is not a separate regulation — it is a structured pre-compliance commitment framework operating in parallel with [the AI Act](/en/glossary/eu-ai-act/) phased applicability.
The Pact is structurally similar to the Climate Change "Race to Zero" voluntary commitment frameworks: signatories make specific public commitments, report progress, and gain reputational positioning while regulatory infrastructure builds out.
## How the AI Pact works
The Pact operates through three commitment layers.
### Pillar 1: General principles
All signatories commit to a set of foundational principles that align with the AI Act spirit:
- Transparency about AI system characteristics
- Risk-based assessment of AI use cases
- Engagement with AI Office implementation work
- Information sharing on AI Act preparation
### Pillar 2: Specific actions
Signatories commit to specific concrete actions ahead of regulatory deadlines:
- **AI literacy programmes** — training employees on AI risks and responsible use
- **AI inventory and classification** — mapping AI systems against AI Act categories
- **Risk-management framework** for high-risk AI use cases
- **Transparency disclosures** — labelling AI-generated content, providing user notices
- **Conformity assessment preparation** for high-risk AI systems
- **Internal AI governance** structures and policies
### Pillar 3: Joint commitments and reporting
Signatories report on Pact commitment implementation through structured reporting cycles. The AI Office maintains an aggregate view of Pact progress and publishes thematic learnings.
## Who signs the AI Pact
As of 2026, the Pact has over 100 signatories spanning:
### Major foundation-model providers
OpenAI, Anthropic, Google, Mistral AI, Aleph Alpha, and others have signed, committing to specific GPAI Code of Practice adherence and broader transparency commitments.
### European AI deployers
Major European enterprises — including in financial services, healthcare, mobility, retail, and public sector — have signed at the deployer level, committing to specific AI governance, risk-management, and transparency practices for their AI use cases.
### European AI startups and SMEs
Growing number of smaller AI businesses signing the Pact, often as part of their go-to-market positioning for enterprise customers requiring AI governance assurance.
### Notable non-signatories
Some AI providers have declined to sign the Pact — typically citing concerns about commitment scope or competitive disadvantage. Non-signatures are themselves a market signal.
## Why the AI Pact matters
### 1. Demonstrates feasibility of AI Act compliance
The Pact serves as a real-world demonstration that AI Act compliance is operationally feasible. Signatories implement requirements ahead of mandatory deadlines, generating practical learnings that inform AI Office implementing acts.
### 2. Procurement signal
Pact signature is increasingly relevant in B2B AI procurement. Enterprise buyers evaluating AI vendors increasingly include Pact-signature status as a positive procurement signal — particularly in regulated industries where AI governance maturity matters.
### 3. Bridge to formal compliance
For organisations not yet directly affected by AI Act mandatory provisions, Pact participation provides a structured framework for building governance capability that will be needed when formal applicability arrives.
### 4. EU AI ecosystem coordination
The Pact creates a structured forum for AI providers, deployers, and the AI Office to discuss implementation challenges, share best practices, and identify regulatory gaps. This is materially valuable for the broader EU AI governance ecosystem.
### 5. Differentiating European AI vendors
European AI providers (Mistral, Aleph Alpha, others) use Pact participation as part of their differentiation against US foundation models. The implicit positioning: European AI vendors aligned with European AI governance, US vendors complying via external requirement only.
## AI Pact vs related frameworks
| Framework | Subject | Status |
|-----------|---------|--------|
| **AI Act** | Mandatory regulation | In force, phased applicability |
| **AI Pact** | Voluntary pre-compliance | Active since 2024 |
| **GPAI Code of Practice** | Voluntary GPAI provider commitments | Published 2025 |
| **AI Liability Directive** | Civil liability for AI harm | Proposed, contested |
| **Member State AI Strategies** | National AI policy frameworks | Per-Member-State |
The Pact complements rather than replaces other frameworks. It is the voluntary pre-compliance layer below mandatory regulation.
## How to participate in the AI Pact
For organisations interested in signing:
### Provider track (AI system developers)
Submit application to the AI Office identifying which AI systems are in scope and committing to specific Pact pillars. Participation includes regular reporting and engagement with AI Office implementation work.
### Deployer track (AI users)
Commit to specific AI governance, risk-management, and transparency practices for AI use cases within your organisation. Particularly relevant for organisations in regulated industries or with significant employee or customer AI exposure.
### SME track
Streamlined participation track for smaller organisations with proportionate commitments and reduced reporting burden.
## What it means in practice
### For AI providers serving EU customers
Pact signature is increasingly procurement-relevant. Non-signature is a discussion point in enterprise sales cycles. For major providers, signature is effectively expected.
### For European AI deployers
Pact participation builds governance capability ahead of mandatory AI Act applicability and signals AI-governance maturity to regulators, customers, and employees.
### For European AI startups
SME-track participation is a strategic positioning opportunity. Pact signature is a credible signal for enterprise sales cycles where AI governance maturity matters.
### For policy and compliance teams
Pact reporting provides structured documentation of AI Act preparation work that supports later formal compliance documentation.
### For procurement teams
Include Pact-signature status in AI vendor evaluation criteria. Non-signature is not necessarily disqualifying but warrants discussion.
## Practical implications
- **For most European businesses considering AI deployment**: Pact framework offers a structured pre-compliance pathway
- **For AI vendors serving EU customers**: Pact participation is increasingly procurement-relevant
- **For European AI startups**: SME-track participation can support enterprise go-to-market positioning
- **For policy watchers**: Pact signature dynamics are a leading indicator of broader AI Act compliance trajectory
- **For investors**: Pact engagement is part of European AI vendor diligence frameworks
The AI Pact is the EU AI governance ecosystem's voluntary pre-compliance bridge. Its uptake and operational learnings shape how the AI Act actually plays out in production through 2026-2027.
Was this helpful?
Thanks for your feedback!