The short answer
“AI Belgium” is not, by itself, the name of one universally recognized company, artificial-intelligence system, or single owner. It is an ambiguous search phrase that can refer to several different things:
- Artificial intelligence in Belgium: the country’s research institutions, startups, companies, public services, and regulatory environment.
- AI 4 Belgium: a national AI strategy and coalition launched in 2019.
- Belgium’s National Convergence Plan for the Development of Artificial Intelligence: a later federal policy plan intended to coordinate Belgium’s development as a “Smart AI Nation.”
- Generative AI Belgium or another private community, event, consultancy, or organization using “AI Belgium” in its name.
- Questions about which AI tools are available in Belgium, who develops them, or how Belgian organizations use them.
Consequently, the answers to “whose AI Belgium,” “who AI Belgium,” “what AI Belgium,” and “which AI Belgium” depend on the intended meaning. Belgium does not have one national AI owned by the Belgian state. Rather, it has a distributed AI ecosystem shaped by federal and regional governments, universities, research centers, businesses, civil-society groups, and European Union rules.
What “AI Belgium” usually means
Artificial intelligence in Belgium includes technologies that perform tasks associated with human intelligence, such as recognizing speech or images, making predictions, generating text, detecting anomalies, optimizing processes, and supporting decisions. These systems may be developed by a Belgian company, operated by a Belgian public authority, hosted by an international provider, or built from open-source software. The location of the user does not establish who owns the underlying model.
The phrase is particularly ambiguous because Belgium is a federal and multilingual country. AI policy and economic support are not managed through one centralized national technology agency. The federal state, Flanders, Wallonia, the Brussels-Capital Region, and the Wallonia-Brussels Federation each have relevant responsibilities, institutions, funding instruments, and innovation programs. The European Commission’s country report describes the Belgian strategy as covering federal and regional actions, including measures for Flanders, Wallonia, Brussels, and the Wallonia-Brussels Federation. Belgium AI Strategy Report - AI Watch - European Commission
A useful distinction is therefore:
| Question | Most accurate interpretation |
|---|---|
| What is AI Belgium? | Usually a broad label for artificial intelligence activity, policy, or community initiatives in Belgium |
| Who owns AI Belgium? | No single answer; it depends on the particular company, project, model, website, or organization meant |
| Who develops AI in Belgium? | Universities, research centers, startups, established firms, public bodies, and international companies operating in Belgium |
| Which AI is used in Belgium? | Many systems, including international commercial tools, open-source models, sector-specific software, and locally developed applications |
| How does Belgium govern AI? | Through European Union law, Belgian implementation measures, federal and regional policy, data-protection rules, and sector-specific regulation |
If the query refers to a particular website or product called “AI Belgium,” its legal owner should be checked in that website’s company information, privacy notice, terms of use, or Belgian enterprise registration rather than inferred from the phrase alone.
AI 4 Belgium and the national policy framework
The most important official use of the name is AI 4 Belgium. Belgium launched this national AI strategy in March 2019. It set out seven broad objectives, including policy support for ethics, regulation, skills, and competencies; support for Belgian AI research and innovation; and measures intended to strengthen adoption and competitiveness. AI 4 Belgium
AI 4 Belgium is better understood as a strategy and coalition framework than as a software product. It does not mean that Belgium owns a single AI model called “AI Belgium.” Its purpose is to organize recommendations and policy priorities around the development and responsible use of AI.
The strategy’s core concerns include:
- Research and innovation: strengthening scientific expertise and the ability to turn research into products and services.
- Skills and education: improving digital literacy, specialist AI skills, and the capacity of workers and organizations to adapt.
- Business adoption: helping companies, including smaller firms, understand where AI can create value and how it can be deployed responsibly.
- Ethics and fundamental rights: addressing discrimination, transparency, accountability, privacy, and human oversight.
- Data and infrastructure: improving access to reliable data, computing capacity, cybersecurity, and digital infrastructure.
- Public-sector use: exploring AI in government while protecting citizens’ rights and maintaining accountable decision-making.
- Coordination: connecting federal and regional initiatives instead of treating Belgium’s separate innovation ecosystems as unrelated.
The strategy reflects a recurring Belgian policy challenge: the country has substantial research, industrial, health, logistics, financial, and public-sector capabilities, but coordination is complicated by the division of powers among multiple governments and by Belgium’s Dutch-, French-, and German-speaking communities.
In 2022, Belgium’s Council of Ministers approved the National Convergence Plan for the Development of Artificial Intelligence. The plan proposes concrete actions intended to make Belgium a “Smart AI Nation.” Its stated themes include trustworthy AI, cybersecurity, competitiveness and attractiveness, and the development of a data-driven economy and society. National convergence plan for the development of artificial ...
The word convergence is significant. The plan is intended to align different policy levels and areas rather than replace every regional initiative with one centrally managed program. In practical terms, a Belgian AI project may receive support or oversight from more than one level of government, depending on its sector, location, funding, data, and intended use.
Who develops and uses AI in Belgium?
There is no single “Belgian AI.” Development is distributed across several groups.
Universities and research centers
Belgian universities and research institutions work on areas such as machine learning, natural-language processing, computer vision, robotics, medical AI, optimization, cybersecurity, and responsible AI. Their work may produce academic research, open-source software, patents, spin-offs, or collaborations with companies and public authorities.
Research ownership can also be divided. A university may own some intellectual property, while a company funds development, licenses the resulting technology, or owns improvements made under a collaboration agreement. An AI system’s nationality and ownership therefore cannot be determined simply by identifying where its researchers are located.
Startups and established companies
Belgian startups and larger companies use AI for customer service, fraud detection, industrial maintenance, recruitment support, supply-chain management, medical research, energy optimization, language services, marketing, and document processing. Some develop their own models; others build applications on top of models supplied by international technology firms.
This distinction matters. A Belgian company may offer an AI service while relying on a model, cloud platform, data center, or application programming interface owned by another company. The Belgian company might own the customer-facing application but not the underlying foundation model. Contracts, licensing terms, training data, and intellectual-property arrangements determine the specific rights.
Government and public services
Belgian public authorities may use algorithmic systems for administrative assistance, forecasting, fraud analysis, document classification, translation, or other operational tasks. Such use raises additional questions about legal authority, data protection, procurement, transparency, human review, and the ability of affected people to challenge decisions.
A public authority using an AI system does not necessarily own it. It may have purchased a licensed service, commissioned a supplier to build a system, or used a shared government platform. Public-sector responsibility also depends on the authority’s role: developer, deployer, purchaser, data controller, or decision-maker.
International providers
People and organizations in Belgium can use AI products developed elsewhere. A user may access a foreign company’s chatbot, image generator, coding assistant, translation system, or enterprise model while physically located in Belgium. The service may be subject to European rules because it is placed on or used in the European market, but that does not make it Belgian-owned.
This is why “which AI is Belgium’s AI?” has no single answer. Belgium participates in a wider European and global market rather than operating one exclusive national platform.
How AI is governed in Belgium
Belgian AI governance operates within the European Union’s legal framework. The EU Artificial Intelligence Act uses a risk-based approach. Different obligations apply depending on the nature and risk of an AI system, including rules concerning prohibited practices, high-risk systems, transparency, and general-purpose AI. The European Commission’s AI Office has responsibilities for implementing, supervising, and enforcing parts of the Act, alongside national authorities and other bodies. Governance and enforcement of the AI Act
The Act is only one part of the legal picture. Depending on the use case, an AI system may also be affected by:
- the General Data Protection Regulation and Belgian data-protection law;
- consumer-protection and product-safety rules;
- employment and anti-discrimination law;
- intellectual-property and copyright rules;
- cybersecurity requirements;
- financial, health-care, education, or public-administration regulation;
- rules governing administrative decisions and public procurement.
The relevant obligations depend on what the system does, who deploys it, what data it processes, and whether it influences people’s access to employment, education, credit, public services, health care, or other important opportunities. Calling a tool “AI Belgium” does not determine its legal classification.
Organizations evaluating an AI system should identify at least four roles:
- Provider – the organization that develops or places the system on the market.
- Deployer – the organization that uses the system in practice.
- Data controller or processor – the party responsible for relevant personal-data processing roles.
- Decision-maker – the person or institution that remains accountable for an outcome affecting an individual.
These roles may belong to different organizations. For example, a Belgian employer could deploy a recruitment tool supplied by an international vendor. The vendor may be responsible for the product, while the employer remains responsible for how the tool is used in recruitment and for complying with applicable employment and data-protection duties.
Which AI tools are available in Belgium?
There are four broad categories:
International general-purpose systems
These include chatbots, image generators, coding assistants, transcription tools, and other services offered by global technology companies. Availability, features, language support, data-handling terms, and age or organizational restrictions can vary by provider and over time.
European and Belgian applications
Belgian and European companies build specialized applications for local industries and languages. A product may focus on Belgian public administration, health care, logistics, manufacturing, legal documents, finance, or multilingual communication without being a general-purpose model.
Open-source and open-weight models
Developers in Belgium can use publicly available models and software, subject to each project’s license and technical limitations. “Open-source,” “open-weight,” and “free to use” are not interchangeable terms: a model may expose its weights without providing all training data or permitting every commercial use.
Internally developed organizational systems
Banks, hospitals, manufacturers, universities, and public bodies may build private AI tools using internal data or connect existing models to controlled databases. These systems may never be publicly available and may not be branded as Belgian AI at all.
Choosing among these categories requires more than asking whether a tool is “Belgian.” A responsible evaluation should consider the system’s purpose, accuracy, language performance, privacy terms, security, integration requirements, cost structure, auditability, accessibility, and legal obligations.
What makes the Belgian context distinctive?
Belgium’s multilingual environment creates both an opportunity and a constraint. AI systems may need to handle Dutch, French, German, and English, sometimes in the same organization or interaction. Performance that is acceptable in English may not transfer equally to regional languages, specialized terminology, dialects, or mixed-language documents.
The country’s federal structure also means that an AI project can have different stakeholders depending on where it is developed and deployed. A health-care application, for example, may involve federal rules, regional services, a hospital network, a university, a software supplier, and European data-protection and AI requirements. A system used by a regional administration may be governed through different institutional channels from one used by a federal department.
Belgium’s policy documents therefore emphasize trustworthy and human-centered AI rather than adoption alone. Trustworthiness requires practical controls: suitable data, testing for errors and bias, cybersecurity, clear responsibility, records of system use, human oversight, and a way to correct or contest harmful outcomes. These controls are especially important where AI supports decisions about people rather than merely automating low-risk administrative work.
How to identify the specific “AI Belgium” meant by a query
When the phrase appears online, identify the object before drawing conclusions:
- Website or domain: inspect the footer, legal notice, privacy policy, and terms.
- Company or nonprofit: look for its official legal name and Belgian enterprise number.
- Government strategy: check whether the reference is to AI 4 Belgium or the National Convergence Plan.
- Community or event: distinguish an informal network from a legal entity or public program.
- AI product: identify the provider, model, hosting arrangement, and license.
- News or social-media post: verify whether “AI Belgium” is a descriptive phrase, campaign name, or organization name.
In short, AI in Belgium belongs to an ecosystem, not to one owner. “AI 4 Belgium” is a national strategy, Belgium’s convergence plan is a coordinating policy framework, and the actual technologies used in the country may be owned by Belgian organizations, European institutions, international companies, public bodies, or open-source communities. The exact answer to “whose AI Belgium?” requires the specific product, website, organization, or policy document behind the search.
Sources
The Belgian Artificial Intelligence Landscape
Belgium approaches artificial intelligence not through a single, centralized bureaucracy, but through a multi-tiered, coordinated network shaped by its federal structure. Understanding how AI functions in Belgium requires looking across three dimensions: the federal coordination bodies, the regional innovation hubs, and the underlying legal framework anchored by the European Union.
At the national level, the federal strategy is facilitated primarily through AI4Belgium, a multi-stakeholder coalition initiated to unite public administrations, industry, research institutions, and civil society. Under the coordination of the Federal Public Service Policy and Support (FPS BOSA), AI4Belgium guides national AI alignment, facilitates responsible adoption, and links local initiatives to the broader European AI ecosystem. Alongside this federal coordination, the Belgian Council of Ministers adopted the National Convergence Plan for Artificial Intelligence to streamline investments, public sector adoption, and ethical standards across the country. National convergence plan for the development of artificial ... Artificial intelligence and emerging technologies - BOSA Belgium AI Strategy Report - AI Watch - European Commission
Because economic development, vocational education, and scientific research are largely devolved competencies under the Belgian state model, the actual implementation of AI research and industrial deployment is split across Belgium’s three administrative regions:
- Flanders: Focuses on deep-tech research, industrial automation, and microelectronics through initiatives such as the Flanders AI Research Program and the world-renowned research institute imec. Belgium AI Strategy Report - AI Watch - European Commission Belgium / AI World
- Wallonia: Operates through programs such as Digital Wallonia, which finances AI readiness and adoption programs (e.g., DigitalWallonia4.ai) aimed at modernizing regional manufacturing, healthcare, and logistics sectors. Belgium AI Strategy Report - AI Watch - European Commission
- Brussels-Capital Region: Anchors services, digital government platforms, ethical AI development, and international collaboration via regional agencies like Innoviris and hub.brussels. Belgium AI Strategy Report - AI Watch - European Commission
Governance: Who Leads and Shapes Belgian AI
Due to Belgium’s complex institutional architecture, AI leadership is shared among multiple public authorities and advisory groups rather than resting in a single ministry.
Federal Level (FPS BOSA)
National Convergence Plan for AI
│
AI4Belgium Coalition
┌───────────────────────┼───────────────────────┐
│ │ │
Flanders Region Wallonia Region Brussels Region
(Flanders AI / imec) (Digital Wallonia AI) (Innoviris / FARI)The Federal Hub: FPS BOSA and AI4Belgium
The operational focal point for federal AI initiatives is the Federal Public Service Policy and Support (FPS BOSA). FPS BOSA leads AI4Belgium and coordinates digital transformation across federal government bodies. AI4Belgium serves as an open platform with thematic working groups covering:
- Ethics and Law: Aligning national policy with European norms and fundamental rights.
- Skills and Education: Addressing the digital talent gap across linguistic boundaries.
- Public Services: Deploying AI within civic administrations to optimize citizen services.
- Health, Industry, and Sustainability: Accelerating domain-specific AI experimentation. Artificial intelligence and emerging technologies - BOSA Belgium AI Strategy Report - AI Watch - European Commission AI 4 Belgium - STIP Compass
Regional Authorities
Each region maintains independent economic development and scientific research portfolios:
- Flanders (VLAIO & Department of Economy, Science and Innovation): The Flemish government funds the Flanders AI Program (Vlaams AI-programma), which splits funding across strategic basic research, industrial test beds, and public literacy. The program unites universities such as KU Leuven, Ghent University, the University of Antwerp, Vrije Universiteit Brussel (VUB), and Hasselt University with imec and Flanders Make. Belgium AI Strategy Report - AI Watch - European Commission Belgium / AI World
- Wallonia (Agence du Numérique): The regional digital agency Agence du Numérique (AdN) drives the DigitalWallonia4.ai strategy, collaborating with universities like UCLouvain, the University of Liège, and the University of Mons, along with research centers such as Multitel and CETIC. Belgium AI Strategy Report - AI Watch - European Commission
- Brussels (Innoviris & Regional Government): Brussels leverages its positioning as both the regional and European capital. A primary pillar is FARI (the AI for the Common Good Institute), a joint research initiative by ULB and VUB, backed by the Brussels-Capital Region and the European recovery fund to serve public interests and urban governance. Belgium AI Strategy Report - AI Watch - European Commission
Key Pillars: What Defines the Belgian AI Ecosystem
Belgium’s AI ecosystem does not attempt to compete directly with global hyperscalers on building foundational large language models from scratch. Instead, it concentrates on high-value niche areas: semiconductor integration, edge computing, trustworthy/ethical AI, healthcare, and industrial computer vision.
| Dimension | Primary Entities | Focus Areas |
|---|---|---|
| Microchip & Edge AI Research | imec (Interuniversity Microelectronics Centre) | Neuromorphic computing, energy-efficient edge processors, sensor-level AI. |
| Industrial Computer Vision & Enterprise Solutions | Companies like Robovision, ML6, Tessares | Applied computer vision in agriculture/manufacturing, enterprise data engineering. |
| Ethical & Common Good AI | FARI Institute, KU Leuven Centre for IT & IP Law (CiTiP) | Responsible AI frameworks, algorithmic transparency, urban public policy modeling. |
| Healthtech & Biotech Analytics | VIB (Flemish Institute for Biotechnology), University Hospitals (UZ Leuven, Erasme, CHU de Liège) | Precision medicine, drug discovery, genomic sequence analysis, automated diagnostic imaging. |
Advanced Hardware and Nanotechnology
A cornerstone of Belgium’s technological weight is imec, based in Leuven. As one of the world's leading research hubs in nanoelectronics, imec designs the physical hardware architecture required for future AI processing. This includes ultra-low-power edge computing devices, photonic chips, and advanced semiconductor packaging that run machine learning models locally on edge devices without relying entirely on remote data centers. Belgium / AI World
Applied Enterprise and Industrial Automation
The private sector ecosystem features specialized systems integrators, applied AI engineering agencies, and business-to-business (B2B) startups. Belgian firms frequently apply predictive maintenance, quality control computer vision, and real-time operations modeling across the manufacturing, pharmaceutical, and logistics corridors of Antwerp, Ghent, Charleroi, and Liège. Belgium / AI World
Legal and Policy Framework: Which Regulations Apply
Operating AI within Belgium is governed by an overlapping hierarchy of regional directives, Belgian national statutory law, and European Union regulations.
┌────────────────────────────────────────────────────────┐
│ EU AI Act (Regulation EU) │
│ Risk-based classification, prohibited practices, │
│ high-risk conformity, transparency obligations │
└───────────────────────────┬────────────────────────────┘
│ Direct Enforcement
┌───────────────────────────┴────────────────────────────┐
│ Belgian Transposition & National Authorities │
│ Data Protection Authority (APD / GBA) │
│ National Market Surveillance Authorities │
│ FPS Economy / BOSA (Digital Governance Convergence) │
└───────────────────────────┬────────────────────────────┘
│ Execution & Oversight
┌───────────────────────────┴────────────────────────────┐
│ Regional Operational Policies & Research Grants │
│ VLAIO (Flanders) · AdN (Wallonia) · Innoviris (Bxl) │
└────────────────────────────────────────────────────────┘The EU Artificial Intelligence Act
As an EU Member State, Belgium does not create a completely detached, standalone domestic compliance system. The foundational rulebook is the EU Artificial Intelligence Act, which follows a tiered risk hierarchy:
- Unacceptable Risk: Banned systems (e.g., social scoring, manipulative dark patterns).
- High Risk: Systems used in critical infrastructure, medical devices, educational access, and employment scoring. These face strict conformity assessments, data quality standards, human oversight protocols, and post-market monitoring.
- Specific Transparency Risk: Systems like conversational agents and deepfakes that require clear labeling to ensure end users know they are interacting with AI.
- Minimal/Low Risk: The majority of common applications (e.g., spam filters, recommender engines), which operate under baseline market rules without heavy certification hurdles. Belgium AI Regulation Overview | Regulations.AI
Data Privacy and the APD/GBA
Belgium’s Data Protection Authority (Autorité de protection des données / Gegevensbeschermingsautoriteit - APD/GBA) plays a direct role in AI regulation. Because the majority of practical AI models ingest or generate personal data, the General Data Protection Regulation (GDPR) acts as an active constraint:
- Automated Individual Decision-Making: Under GDPR Article 22, individuals retain the right not to be subject to solely automated decisions that produce legal or similarly significant effects, mandating meaningful human intervention.
- Data Minimization and Fairness: AI training datasets must respect purpose limitation, data accuracy, and user rights to deletion or rectification where applicable.
- Impact Assessments: Any high-risk machine learning system processing personal data at scale requires a formal Data Protection Impact Assessment (DPIA) before deployment in Belgian jurisdictions.
Strategic Opportunities and Structural Challenges
While Belgium possesses institutional strengths, its artificial intelligence ecosystem faces distinct challenges arising from geography, funding size, and constitutional organization.
Advantages and Opportunities
- Dense Academic and Research Network: Belgium has a high concentration of research institutions per square kilometer. Interdisciplinary centers link engineering departments directly with humanities and legal clinics (such as KU Leuven's CiTiP), fostering reliable AI compliance by design. Belgium / AI World
- Central European Location and Logistics Infrastructure: The presence of European institutions in Brussels alongside the Port of Antwerp-Bruges provides testing environments for supply chain intelligence, customs automation, and multilingual public governance. Belgium AI Strategy Report - AI Watch - European Commission
- Semiconductor Leadership: The global dependence on imec’s lithography and chip-architecture research guarantees Belgium an enduring voice in the hardware foundations of artificial intelligence. Belgium / AI World
Bottlenecks and Structural Limits
- Venture Capital and Scaling Capacity: Belgian startups frequently face capital constraints when attempting to scale from Series A to later growth stages. High-growth firms often seek growth capital from the United Kingdom, the United States, or larger European economies, risking talent or intellectual property migration.
- Institutional Fragmentation: Because competencies are divided between the federal state, three geographic regions, and three linguistic communities, initiatives risk overlapping efforts or siloed procurement standards. The National Convergence Plan was designed specifically to counter this fragmentation, yet regional initiatives often maintain distinct funding pools and administrative requirements. National convergence plan for the development of artificial ... Belgium AI Regulation Overview | Regulations.AI
- Talent Retention: With proximity to major tech hubs in Paris, Amsterdam, London, and Berlin, Belgian firms face ongoing competition to attract and retain specialized machine learning engineers and data scientists against larger private-sector compensation packages. Belgium AI Strategy Report - AI Watch - European Commission
Navigating the Belgian AI Landscape
For organizations looking to engage with the Belgian AI ecosystem, navigation depends on the operational goal:
- For Technical Collaboration & Microelectronics: Connect with imec or academic consortia through the Flanders AI Research Program. Belgium AI Strategy Report - AI Watch - European Commission Belgium / AI World
- For Regional Business Grants & Implementation: Flemish organizations route through VLAIO; Walloon organizations operate through Digital Wallonia (DigitalWallonia4.ai); Brussels-based organizations work with Innoviris and hub.brussels. Belgium AI Strategy Report - AI Watch - European Commission
- For Policy, Civic Tech, and Multilateral Standards: Engage with FPS BOSA, the AI4Belgium working groups, or the FARI Institute in Brussels for public sector use cases and civic applications. Artificial intelligence and emerging technologies - BOSA Belgium AI Strategy Report - AI Watch - European Commission
Sources
- [1]National convergence plan for the development of artificial ...bosa.belgium.be
- [2]Artificial intelligence and emerging technologies - BOSAbosa.belgium.be
- [3]Belgium AI Strategy Report - AI Watch - European Commissionai-watch.ec.europa.eu
- [4]Belgium / AI Worldaiworld.eu
- [5]AI 4 Belgium - STIP Compassstip.oecd.org
- [6]Belgium AI Regulation Overview | Regulations.AIregulations.ai
I'll research Belgium's AI landscape and current governance status before writing.Now let me check Belgium's AI research and compute ecosystem plus the latest EU-level timing changes.One more check on Belgian AI adoption and notable companies.## Reading the question: a cluster of related queries about AI in Belgium
The phrase "how ai belgium," along with the variants "who," "whose," "what," and "which ai belgium," is not a single well-formed question. Read together, these fragments point at one practical subject: how artificial intelligence is governed, funded, built, and used in Belgium — and who is responsible for each part of that. This article answers that composite question. If your intent was narrower (for example, a specific Belgian company whose name resembles "AI Belgium," or the AI4Belgium coalition in particular), the relevant sections below should still point you to the right actor.
The short answer: Belgium does not have its own AI law. AI is regulated in Belgium by the EU Artificial Intelligence Act (Regulation (EU) 2024/1689), which applies directly without national transposition. What Belgium owes at national level is enforcement machinery — designated authorities, powers, and penalties — and that machinery is still incomplete. The federal coalition agreement names the telecom regulator BIPT/IBPT as the intended lead market-surveillance authority, with the FPS Economy coordinating, but at the time of writing this designation remains government policy rather than enacted statute. Belgium missed the EU's 2 August 2025 designation deadline, and reporting in mid-2026 described the practical consequence bluntly: the obligations bind companies, but the supervisory office that would enforce them does not yet formally exist, leaving the courts as the enforcement route.
What rules actually apply: the EU layer
Because the AI Act is a regulation rather than a directive, Belgian firms have been bound by its substantive provisions on the EU timetable regardless of domestic legislative progress. The prohibited-practice rules in Article 5 have been enforceable since 2 February 2025, and general-purpose AI (GPAI) model obligations since 2 August 2025. The headline change since then is the Digital Omnibus on AI (Regulation (EU) 2026/1744), published in the Official Journal on 24 July 2026 and in force from 27 July 2026, which deferred the core high-risk regime.
| Obligation | Original date | Current date |
|---|---|---|
| Prohibited practices (Art. 5) | 2 Feb 2025 | unchanged |
| GPAI model obligations (Art. 53) | 2 Aug 2025 | unchanged |
| Art. 50 transparency (chatbot disclosure, deepfake labelling) | 2 Aug 2026 | unchanged |
| Art. 50(2) machine-readable marking, systems already on market | 2 Aug 2026 | 2 Dec 2026 (grace period) |
| Stand-alone high-risk systems (Annex III) | 2 Aug 2026 | 2 Dec 2027 |
| High-risk AI embedded in regulated products (Annex I) | 2 Aug 2027 | 2 Aug 2028 |
| Member-state obligation to run at least one AI sandbox | 2 Aug 2026 | 2 Aug 2027 |
| Systems intended for use by public authorities | 2 Aug 2026 | 2 Aug 2030 |
The deferral is partial and easy to misread. Transparency duties, GPAI requirements, the EU AI Office's powers, and governance obligations stayed on the August 2026 schedule; only the Chapter III high-risk regime and a narrow set of adjacent dates moved. The Omnibus also added a new prohibited practice covering AI systems used to generate non-consensual sexual or intimate content, including child sexual abuse material, applying from 2 December 2026. The common reading among practitioners is that the delay reflects unfinished technical standards rather than a softening of expectations.
Who supervises AI in Belgium
The AI Act requires each member state to designate national competent authorities, including at least one notifying authority (which evaluates and notifies conformity assessment bodies) and at least one market surveillance authority. Any designated authority can issue implementation guidance, request documentation from providers, and require providers of high-risk systems to demonstrate compliance. Belgium has not completed these statutory designations; trackers monitoring national implementation still record the Article 70 and Article 74 roles as pending, with BIPT described as announced rather than appointed.
That said, "no lead authority yet" does not mean "no supervision." Several Belgian regulators already have jurisdiction over AI through sector mandates or the GDPR, and the AI Act itself channels certain system types to existing supervisors:
- Financial and insurance AI — credit scoring, underwriting, and similar systems fall to the FSMA and the National Bank of Belgium, which act as market surveillance authorities for the sector.
- Biometrics, law enforcement, justice — supervised by the Belgian Data Protection Authority (APD/GBA), which also published guidance on AI and the GDPR in September 2024 and has positioned itself as a central AI supervisor.
- Anything processing personal data — the GDPR applies in parallel, independently of AI Act classification.
- Fundamental-rights oversight under Article 77 — Belgium's draft arrangement spreads this across a long list of sectoral bodies, reported as roughly twenty-one authorities.
- Everything else — intended to land with BIPT once legislation passes.
Penalty exposure comes from the Regulation itself: up to €35 million or 7% of worldwide annual turnover for prohibited practices, up to €15 million or 3% for most other breaches, and up to €7.5 million or 1% for supplying incorrect information, with the lower of the two figures applying to SMEs. Belgium is still adopting the national instrument that sets out procedures, powers, and domestic penalty rules. Anyone making compliance decisions on the strength of the current gap should take qualified Belgian legal advice; the enforcement picture is unsettled and changes with each legislative step.
How Belgium steers AI: soft law and a federated strategy
Belgium's distinctive feature is not its regulation but its governance geometry. Competences for research, economy, education, and innovation are split between the federal state and the Flemish, Walloon, and Brussels-Capital levels, so AI policy is assembled from coordinated plans rather than a single statute. The approach is predominantly policy-led and non-coercive, favouring grants, coaching programmes, self-assessment tools, and test environments over domestic prohibitions.
The principal instruments are:
- AI4Belgium — launched in March 2019 as the national AI coalition, coordinated by the federal public service FPS BOSA. It is a community-driven network rather than an agency: its objectives include policy support on ethics and skills, mapping the Belgian AI landscape, animating the national community, channelling EU funding, and driving industrial uptake. With an annual budget under €1 million, its influence lies in consensus-building and in supplying input to formal policy. It also maintains practical artefacts such as an ALTAI-based trustworthiness assessment tool and guidance on responsible public procurement and generative AI in government.
- The National Convergence Plan for the Development of AI — approved by the Council of Ministers on 28 October 2022, built around the ambition of a "#SmartAINation" and structured in multiple pillars including ethical and responsible AI. A multi-stakeholder steering committee coordinates it across federal, regional, and community levels. The AI4Belgium roadmap, which had recommended at least €1 billion of AI investment by 2030, directly informed it.
- Regional programmes — the Flanders AI Policy Plan (reported at roughly €32 million per year, split across business uptake, research, and training/ethics), Wallonia's DigitalWallonia4.ai with the ARIAC research programme, and Brussels' innovation agency Innoviris, alongside FARI, the AI institute created under Brussels' AI policy.
Which institutions build AI — research, compute, and sovereignty
On the research side, Belgium's capability is concentrated in a handful of recognisable actors. The Flanders AI Research Program (FAIR), now in a second cycle running 2024–2028, is a strategic basic research consortium of eleven partners: the five Flemish universities (KU Leuven, Ghent, Antwerp, Hasselt, VUB) and six research centres (imec, Flanders Make, VIB, VITO, Sirris, ILVO). Wallonia's counterpart ecosystem includes the TRusted AI Labs (TRAIL), with firm-facing programmes such as Start IA and Tremplin IA supporting proofs of concept. imec, founded in 1984, anchors the hardware end — semiconductors, edge-AI hardware, and quantum — and Wallonia's Lucia supercomputer at Cenaero adds regional GPU capacity.
Compute is where the "whose AI" question becomes concrete. Belgium has no frontier-model laboratory of its own and has pursued access to European infrastructure rather than building national-scale training capacity. In 2025 the federal digital minister, with regional support, applied for a full AI Factory under the InvestAI programme, proposing two interconnected sites — the Green Energy Park in Zellik and a site in Charleroi near Cenaero — at an estimated €80 million, half of it EU-subsidised. In October 2025, EuroHPC selected six new AI Factories and thirteen antennas; Belgium, which had bid for both, was awarded an AI Factory Antenna. The resulting BE-AIFA project runs 2026–2029 with a €10 million budget co-funded equally by EuroHPC and Belgian federal and regional governments, coordinated by imec across a consortium of twenty-three universities, research centres, and public institutions from all three regions and the federal level. It gives Belgian organisations brokered access to systems such as LUMI in Finland and JUPITER in Germany, plus dataset access, expert services, and training. Data-sharing capacity is being developed in parallel through initiatives such as the Flemish data utility Athumi and the Flemish Data Strategy.
The pattern is consistent: Belgium's strategy is to be an excellent consumer and integrator of AI, and a specialist contributor in semiconductors and applied research, rather than a builder of large general-purpose models.
How widely AI is actually used
Adoption data supports that framing. In Eurostat's enterprise ICT survey for reference year 2025, 34.5% of Belgian enterprises with ten or more employees used at least one AI technology, placing Belgium fourth in the EU behind Denmark (42%), Finland (about 38%), and Sweden (35%), against an EU27 average of roughly 20%. The trajectory is steep: 10.3% in 2021, 13.8% in 2023, 24.7% in 2024, 34.5% in 2025 — moving Belgium from seventh to fourth place. Adoption is strongly size-dependent, at about 28.8% for firms with 10–49 staff, 54.5% for 50–249, and 76.4% for 250 or more, the last of which ranks second in the EU. Statbel collects the Belgian component of the survey, and the indicator measures only whether any tracked AI technology is in use, not depth of deployment.
Practical implications and limits
For an organisation operating in Belgium, three things follow. First, the absence of a Belgian AI statute is not a compliance holiday — Article 5, GPAI, and Article 50 transparency duties already bite, and the deferred high-risk deadlines (December 2027 and August 2028) assume preparation begins long before. Second, your effective supervisor may already be identifiable: a bank or insurer looks to the FSMA and National Bank, and any personal-data processing keeps you inside the APD/GBA's remit regardless of how BIPT's designation is finalised. Third, support is regional before it is federal — the funding, coaching, and sandbox-style assistance most firms can actually use comes through VLAIO and the Flemish plan, DigitalWallonia4.ai, or Innoviris and FARI, with BE-AIFA as the route to serious compute.
Two caveats are worth stating plainly. Belgium's institutional picture is in motion, so any statement about designated authorities, pending bills, or sandbox availability should be re-verified against current official sources before it is relied on. And because competence is genuinely split across levels of government, generic "Belgian AI policy" summaries can mislead: which rules, which funding, and which regulator apply depends on your sector, your region, and whether you are a provider or a deployer.