BgGPT instead of ChatGPT: Artificial Intelligence in Bulgaria – Between ambitious vision and structural gap

Bulgaria as Europe's secret AI lab? What German companies need to know now

Excellent research, (still) weak economy: The harsh AI reality in the Balkans

When people think of the pioneers of artificial intelligence in Europe, they usually look to Paris, London, or Berlin – Bulgaria is rarely on their radar. But a closer look reveals a fascinating paradox: On the one hand, the EU country boasts state-of-the-art supercomputers, internationally renowned research institutes like INSAIT, and even its own open-source language model called BgGPT. On the other hand, the Bulgarian business landscape still lags behind the rest of Europe in terms of digital competence and AI adoption.

This very gap between visionary cutting-edge research and a hesitant, analog business reality holds enormous economic potential. Particularly against the backdrop of the planned introduction of the euro in 2026, massive EU funding, and the stringent requirements of the new EU AI Act, the Bulgarian market is gaining momentum. A strategic window of opportunity is opening for European tech companies and providers of data protection-compliant, secure AI solutions. This article examines the asymmetry of the Bulgarian AI ecosystem, analyzes the regulatory challenges, and pinpoints where structural need meets genuine willingness to buy.

Why an EU member state with a supercomputer and its own language model still hasn't arrived in the AI ​​age

Bulgaria is not a country one would instinctively associate with a pioneering role in artificial intelligence. Yet current figures and developments paint a picture that is considerably more nuanced than initial impressions suggest: on the one hand, remarkable research infrastructure and strategic ambitions at the state level; on the other, a large portion of the population and a business sector that lags far behind the EU average in terms of digital maturity. This tension between strategic vision and lived reality shapes the current state of AI development in the country – and simultaneously determines where real market opportunities lie for foreign providers.

Government strategy: The framework exists, but implementation lags behind

The formal basis for Bulgaria's AI policy is the concept paper on the development of artificial intelligence until 2030, adopted by the Ministry of Transport, Information Technology and Communications in 2020. This document defines six strategic pillars: infrastructure, education, research, data use, sectoral innovation, and ethical AI development. On paper, this is a coherent and ambitious agenda.

The problem lies in the implementation. In its assessment of EU member states within the framework of the coordinated AI plan, the OECD noted that while the Bulgarian strategy formulates a comprehensive vision, it lacks a concrete action plan with clear implementation steps and timelines. Bulgaria thus shares a fate familiar to many Eastern European EU members: the strategic document serves primarily as a signal to Brussels, rather than as an internal steering instrument.

The Ministry of E-Government, which has operated independently for several years, has nevertheless taken concrete steps. Within the framework of the National Action Plan 2022–2024, standards for the use of AI in the public sector were developed, with a particular focus on human rights issues, algorithmic decision-making, and transparency. This initiative is noteworthy in that it demonstrates that at least parts of the Bulgarian administration take the ethical dimension of AI seriously – even though the practical implementation of these standards is still pending.

In parallel, Bulgaria is investing significant sums in digital transformation through its Recovery and Resilience Plan. According to the European Commission, the country has earmarked a total budget of €2.2 billion for digital objectives – equivalent to approximately 2.3 percent of its GDP. Part of this funding is being allocated to expanding supercomputing capacity: the Discoverer high-performance computer, operating as part of the European EuroHPC project, and the Avitohol system, currently under development, underscore Bulgaria's focus on the infrastructure side of AI. With the selection of Bulgaria as the location for one of six new European AI factories – a €90 million project supported by INSAIT and the Sofia Tech Park – the country reached another structural milestone in March 2025.

Research excellence as an asymmetric advantage

The most striking and unexpected feature of the Bulgarian AI landscape is the level of academic research. The Institute for Computer Science, Artificial Intelligence and Technology (INSAIT) at Sofia University, founded in 2022 in cooperation with ETH Zurich and EPFL, has quickly become an internationally recognized research institution. It is the first Eastern European institution comparable to leading Western universities in terms of research conditions and international collaboration.

The numbers speak for themselves: 14 research papers were accepted at the International Conference on Computer Vision (ICCV) 2025, placing INSAIT on a par with representatives from Google, Meta, and Sony. Seven more papers were accepted at CVPR 2025 in Nashville. The institute has mobilized over $100 million in funding and partnerships with organizations such as Google DeepMind and AWS.

The most significant product of this research is BgGPT, Bulgaria's first in-house open-source language model. The current version, BgGPT 1.0, is based on Google's Gemma 2 models and has been trained on over 100 billion Bulgarian and English tokens. Released under an Apache 2.0 license, the model is freely available for use by both public institutions and businesses and can be run entirely locally – without transferring data to external servers. In Bulgarian language tasks, it significantly outperforms larger models such as Mixtral-8x7B. This is not an academic curiosity, but a practically relevant tool explicitly designed for use in education, healthcare, government, and business.

This high level of research combined with a massive need to catch up across the business sector is the characteristic asymmetry of the Bulgarian AI ecosystem: an excellent top without a broad base.

Corporate sector: a laggard in Europe, but with growing islands

The most sobering finding from the EU comparison is this: Only 29.3 percent of Bulgarian companies use advanced digital technologies – such as cloud computing, data analytics, or AI – placing Bulgaria last among all EU member states. The EU average is 54.6 percent. Older data shows even more precisely that only three percent of Bulgarian companies actively used AI. Current figures from 2025 indicate that approximately 6.5 percent of companies have integrated AI into their operations.

This reluctance has several structural causes. First, there is a shortage of qualified personnel: only 4.6 percent of the workforce are ICT specialists, slightly below the EU average of 5 percent. Second, graduates' ability to meet the demands of the labor market is low – the QS Future of Work report shows a skills fit score of only 37.6 for Bulgaria. Third, awareness of the productive use of AI is still rudimentary in many small and medium-sized enterprises.

However, within this overall weak adoption rate, there are areas that are significantly more dynamic. The financial sector – particularly fintechs and digital payment services – is among the early and comparatively mature adopters of AI technologies. The startup Payhawk, Bulgaria's first unicorn, exemplifies this development. In the logistics sector, Dronamics, a cargo drone operator with over €52 million in venture capital funding, uses AI for autonomous flight route selection. In the healthcare sector, Sensika develops AI-supported diagnostic tools based on smartphone cameras, which are now used in over 25 countries.

The startup ecosystem as a whole comprised around 90 AI product companies in 2024, which together raised just under €54.7 million in funding – less than in the record year of 2023, which reached €101 million. However, the base remains thin behind the top performers, and the quality of newly founded companies is not growing at the same pace as the available funding.

Digital divide: Private users caught between curiosity and lack of skills

The societal aspect of AI development in Bulgaria is characterized by a pronounced digital divide. Only 35.5 percent of the population possesses basic digital skills, compared to an EU average of 55.6 percent – ​​placing Bulgaria 26th out of 27 EU member states. A mere eight percent of the working-age population has digital skills above a basic level.

This data is crucial for understanding AI use in the private sphere. While younger, urban Bulgarians readily use Western AI tools like ChatGPT or Gemini, access remains limited for older, rural, or less educated populations. The language barrier plays a particularly significant role here: English proficiency is not widespread outside of cities and university circles, which considerably restricts the use of English-language AI tools. Therefore, the development of BgGPT as a Bulgarian-language model is not merely of academic interest, but addresses a real obstacle to access.

The Bulgarian Ministry of Labor and Social Policy has announced training programs for over 660,000 Bulgarians by 2026, financed through the National Recovery and Resilience Plan and the Human Resource Development Program. The goal is to raise the population's digital literacy across the board. Whether these measures will be sufficient to close the structural deficits within a relevant timeframe remains to be seen – the need for learning is immense, and the quality of implementation of public digital programs in Bulgaria has historically been inconsistent.

Public authorities and administration: the desire for reform and the bureaucratic weight

Bulgaria's public administration is in a mixed state. On the one hand, Bulgaria achieves a score of 91.9 out of 100 in the digitalization of public services for businesses, exceeding the EU average. This means the country is well-positioned in the formal provision of digital administrative services for companies. On the other hand, German and international investors regularly complain about a slow, sometimes corruption-prone bureaucracy whose digital interface does not align with its analog internal processes.

The use of AI within public administration itself remains cautious. The Ministry of E-Government is working on a framework for algorithmic decision-making in social services, employment agencies, and law enforcement. Sensitive applications—such as the automated distribution of social benefits or risk assessment in cases of domestic violence—are a focus because they touch upon fundamental rights issues. Accordingly, implementation is proceeding cautiously and with the intensive involvement of civil society actors.

However, the e-government sector offers genuine potential for AI-driven efficiency gains. With one of the highest levels of online service digitization for businesses in the EU, the foundation exists for further automation – for example, in document processing, application handling, or internal administrative knowledge management. Municipalities and city authorities in Sofia and Plovdiv are showing increasing interest in intelligent solutions for traffic management, citizen services, and resource planning. This is an area where external solution providers can find real entry opportunities.

A new dimension of digital transformation with 'Managed AI' (Artificial Intelligence) – Platform & B2B solution | Xpert Consulting - Image: Xpert.Digital

Here you will learn how your company can implement customized AI solutions quickly, securely and without high entry barriers.

A managed AI platform is your all-inclusive, worry-free solution for artificial intelligence. Instead of dealing with complex technology, expensive infrastructure, and lengthy development processes, you receive a ready-made solution tailored to your needs from a specialized partner – often within just a few days.

The key advantages at a glance:

⚡ Rapid implementation: From idea to ready-to-use application in days, not months. We deliver practical solutions that create immediate added value.

🔒 Maximum data security: Your sensitive data stays with you. We guarantee secure and compliant processing without sharing data with third parties.

💸 No financial risk: You only pay for results. High upfront investments in hardware, software, or personnel are completely eliminated.

🎯 Focus on your core business: Concentrate on what you do best. We take care of the entire technical implementation, operation, and maintenance of your AI solution.

📈 Future-proof & scalable: Your AI grows with you. We ensure continuous optimization and scalability, and flexibly adapt the models to new requirements.

More information here:

• The Managed AI Solution - Industrial AI Services: The Key to Competitiveness in the Services, Industry and Mechanical Engineering Sectors

GDPR and AI regulation: Formally compliant, practically under-regulated

A common misconception about Bulgaria is that data protection is taken less seriously there than in Germany. This is formally incorrect: As an EU member state, the GDPR has been directly applicable law since May 25, 2018, supplemented by the Bulgarian Personal Data Protection Act (PDPA), which was amended accordingly in 2019. The Commission for Personal Data Protection (CPDP) is the responsible supervisory authority. However, legal experts describe the CPDP's approach as reactive – proceedings are predominantly initiated based on complaints, not through proactive monitoring campaigns.

The CPDP's 2024 activity report shows that complaints and reports in sectors such as electronic communications, online gambling, and direct marketing predominate. AI-specific proceedings are still rare – not because AI is of little relevance to data protection in Bulgaria, but because the authority itself has not yet developed an active AI oversight strategy. The level of GDPR enforcement is significantly lower than in Germany, France, or the Netherlands, where data protection authorities have already imposed substantial fines on AI providers such as OpenAI.

The EU AI Act, which entered into force in August 2024 and will be fully applicable by August 2026, fundamentally changes this picture. As an EU member, Bulgaria is fully subject to the risk classification regime of the AI ​​Act. The prohibitions on certain AI practices have been in effect since February 2025. This means that companies operating in Bulgaria or with Bulgarian customers are subject to the same compliance obligations as anywhere else in the EU – regardless of whether the national supervisory authority is already actively enforcing them.

The legal framework for AI in Bulgaria – including liability issues, contract law, and intellectual property law – is fully aligned with EU law. However, national implementing regulations and the designation of specific AI supervisory authorities, which the AI ​​Act mandates by August 2026, are still lacking. This regulatory gap creates a degree of tolerance for experimental AI implementations in the short term, but in the long term, it represents a risk factor.

A key characteristic of the Bulgarian market is that GDPR compliance is legally mandatory and is rigorously enforced in regulated sectors (banking, healthcare, public administration), but an active data protection culture in the sense of a proactive privacy-by-design approach is significantly less pronounced than in Germany. Companies in sensitive sectors therefore operate within a formally identical legal framework, but one that is enforced differently in practice.

Data sovereignty as a market gap: What European privacy-first architectures offer

Precisely because the Bulgarian business environment, with its rapidly growing middle class, is to be introduced to AI productivity tools, an interesting gap is emerging in the market: Many companies want to use AI, but with sensitive data – be it customer data in the financial sector, patient information in healthcare, or personal records in public administration. At the same time, they often lack the internal expertise to independently implement secure and GDPR-compliant solutions.

Here, architectural approaches are gaining importance that view data protection not as an afterthought compliance feature, but as a structural design principle. This approach doesn't simply involve choosing an EU-based cloud provider, but rather a more fundamental security architecture: Sensitive data – personal information, client data, patient data, confidential trade secrets – is identified, masked, or anonymized at the input level before being forwarded to external AI models. The result is a risk distribution that differs fundamentally from the simple SaaS usage of American hyperscalers: The company can utilize powerful external language models without any sensitive data leaving its own infrastructure boundaries in a readable form.

This approach is not new – it has become the de facto standard in regulated industries in Germany and Western Europe. However, it is still largely unknown in the Bulgarian market. Banks, insurance companies, hospitals, and public institutions that want to implement AI face a choice between the uncontrolled use of consumer AI tools – often initiated by individual employees without organizational oversight – and completely foregoing the productivity gains that AI can offer.

European AI platforms with a privacy-first architecture offer a concrete solution to this dilemma. Fully on-premises options, including air-gapped deployments for high-security environments, enable regulated organizations to run AI within their own infrastructure. Hosting either in EU-certified data centers or entirely in the company's own data center structurally eliminates data sovereignty risks. For institutions that are prohibited from transferring data to external infrastructures under any circumstances—state-owned banks, public administrations, and municipally owned hospitals—this is not an optional convenience, but a mandatory operational requirement.

Furthermore, administrative functions play a crucial role: Single sign-on, role-based access control, full interaction logging, and multi-tenant user management are not merely technical features, but fundamental prerequisites for auditable AI use that meets internal compliance requirements and withstands external audits. In an environment preparing for the EU AI Act and its documentation obligations, this aspect should not be underestimated.

Market opportunities for German suppliers: Where structural demand meets willingness to buy

The introduction of the euro on January 1, 2026, represents a significant economic policy shift that will considerably increase Bulgaria's strategic importance as a market for German companies. Exchange rate risks will be eliminated, transaction costs will decrease, and the already intensive trade relationship – Germany is Bulgaria's most important trading partner with a volume of approximately €12 to €12.4 billion – will be further strengthened. Full Schengen accession in 2025 will also simplify logistics and passenger transport.

This results in the following specific market opportunities for the AI ​​and tech sector:

The most immediate point of contact lies in the area of ​​managed AI for businesses. Bulgarian SMEs and mid-sized companies want to use AI but lack the infrastructure and personnel for independent implementations. Fully managed AI platforms—ideally GDPR-compliant, hosted in the EU, with local language support and clear ISO 27001 certification—directly address this need. Since Bulgarian companies are either unwilling or unable to build their own AI governance infrastructure, managed solutions with clear compliance guarantees are particularly attractive.

In the public sector and at local authorities, automation solutions for document processing, internal knowledge management, and citizen communication are in demand. While Bulgaria has a good rate of digitization for external e-government services, internal administrative processes are often still paper-based or semi-digital. German providers with experience in public compliance requirements—such as the AI ​​Act, GDPR, and information security—who can package these into a turnkey solution, will find a receptive market environment here.

In regulated industries—banking, insurance, healthcare—the need for controllable, auditable AI is particularly pronounced. These sectors cannot afford informal AI use by employees on consumer platforms due to regulatory and liability concerns. They require solutions that offer enterprise-level access control, logging, and data isolation. Germany holds a clear competitive position here: providers with GDPR compliance by design, German server locations or on-premises options, and EU AI Act competence speak a language of trust that American hyperscaler offerings cannot replicate.

Nearshoring IT development for AI products presents another opportunity. Bulgaria boasts a pool of qualified IT professionals at significantly lower labor costs than Germany. Several well-known German companies have already opened IT development centers in Sofia and other Bulgarian cities. Combined with INSAIT's AI research strength, this offers Bulgaria the possibility of utilizing nearshoring capacities specifically for AI product development.

However, structural obstacles must be considered: a shortage of skilled workers despite a favorable starting position – qualified specialists often leave Bulgaria for Western Europe – a bureaucratic administration that can delay project approvals, and a perception of corruption that sometimes overshadows investment decisions. Successful market participants from Germany recommend building partnerships with local stakeholders and utilizing the German-Bulgarian Chamber of Commerce (AHK) networks as a first step.

Sectoral priorities in the AI ​​ecosystem

The current Bulgarian AI ecosystem is clearly concentrated in specific verticals. Natural Language Processing has historically been the strongest domain, which can be explained by the academic environment and the need for Bulgarian-language solutions. Predictive Analytics and Data Science – particularly in the financial sector – are also well-developed.

New areas added in recent years include autonomous systems and drone technology (Dronamics), agritech (Smart Farm Robotix uses AI for precision agriculture), digital identity and trust (Evrotrust), and medical image analysis. The vertical profile of the ecosystem reflects the country's strengths: a strong tradition of mathematical and scientific education, comparatively low labor costs for software developers, and a growing diaspora of Bulgarians in international tech companies, bringing back expertise.

Networking within the ecosystem is still limited. Sofia Tech Park is developing into a central hub, and INSAIT acts as an academic anchor, attracting international talent – ​​4,000 applicants from 150 countries for the Summer AI Program 2025 demonstrate its global reach. However, the bridge between cutting-edge academic research and broad entrepreneurial application remains narrow.

Bulgaria's AI development in the EU context: Opportunities in the catch-up process

Within the EU context, Bulgaria finds itself in a position aptly described as "structural lag with strategic opportunity." Its lag in adoption and digital skills is real, but it also presents a starting point for potential leaps in development: countries that are late to digitize can build directly on current standards without having to rebuild legacy systems. Experience from East Asia demonstrates how quickly laggards can catch up with the right political will and external support.

The strategic situation is favorable: EU membership creates legal certainty and access to funding, the introduction of the euro reduces market entry costs for Western European providers, an ambitious research institute (INSAIT) attracts international talent and capital, and a growing startup scene provides fertile ground for entrepreneurial dynamism.

The critical variable is the interface between strategy and implementation. Where Bulgaria lags most behind – in concrete action plans, systematic training programs, and consistent regulatory enforcement – ​​are precisely those areas where external expertise and proven solutions can have the greatest impact. For German providers offering GDPR-compliant AI solutions designed for European regulations, Bulgaria is therefore not a niche market, but a structurally attractive growth market with medium-term potential.

The key to market success lies not in undifferentiated technology transfer, but in the ability to understand the specific requirements of Bulgaria's regulated industries and public sector: limited internal IT know-how, growing awareness of compliance risks, high demand for managed services, and the desire for reliable, auditable AI solutions that operate within the familiar European legal framework.

The quasi-in-house solution: How Xpert.Digital closes operational gaps in B2B marketing and sales – Smart Content-Driven Business - Image: Xpert.Digital

Xpert.Digital is a data-driven B2B industry hub led by Konrad Wolfenstein . The company acts as an external, quasi-in-house solution for industrial partners, closing operational gaps in marketing, content, and sales – without requiring additional resources on the client side.

More information here:

• The quasi-in-house solution: How Xpert.Digital closes operational gaps in B2B marketing and sales – Smart Content-Driven Business

☑️ Our business language is English or German

☑️ NEW: Correspondence in your native language!

Konrad Wolfenstein

I and my team are happy to be available to you as your personal advisor.

You can contact me by filling out the contact form here simply call me at +49 7348 4088 965. My email address is [email protected]:or

I'm looking forward to our joint project.

☑️ SME support in strategy, consulting, planning and implementation

☑️ Creation or realignment of the digital strategy and digitization

☑️ Expansion and optimization of international sales processes

☑️ Global & Digital B2B trading platforms

☑️ Pioneer Business Development / Marketing / PR / Trade Fairs