A Development Economics Section of the Hungarian Economic Association (MKT) held a roundtable on the economic and security impacts of corporate artificial intelligence (AI) adoption. Participants included Lits Benedek (PhD student, Budapest Corvinus University), Kasznár Attila (Deputy Dean for International Affairs and Development, Department of International Economics, John von Neumann University) and Csáki Csaba (Dean responsible for AI, Budapest Corvinus University). The discussion was moderated by Trautmann László, chair of the MKT Development Economics Section.
European study motivating the discussion and key findings
The event was prompted by a working paper being prepared by the Hungarian Chamber of Commerce and Industry at the request of Eurochambres; the professional part of the study is being developed by Lits Benedek together with Brussels-based experts Cornelius Knaack and Ilaria Colajanni. Preliminary results from the still-ongoing research show pronounced heterogeneity in company-level AI use: small and medium-sized enterprises lag substantially behind larger firms, and most countries have not meaningfully caught up since the first data collection in 2021. The gap between North/West Europe and Southern/Eastern Europe, including Hungary, has remained persistent.
A panel regression in the study finds that a one percentage point increase in AI adoption raises firms’ revenue per employee (labour productivity) by 0.28 percentage points. Among technological uses, process automation, image recognition and marketing/sales applications showed statistically significant associations with productivity gains.
No clear, strong relationship was found between AI adoption and overall employment. There was some evidence of reduced employment among those under 25 associated with adoption, but this effect disappears when examining the broader under-40 group. The extent to which AI is complementary to or substitutive for labour requires further research.
Poland was cited as an interesting case, where strong economic growth occurred in recent years despite relatively low AI adoption. Denmark stands out as a leading adopter overall, and by sector the pharmaceutical industry—after ICT sectors—showed the highest level of AI adoption.
Digitalisation gap and corporate security risks in Hungary
Kasznár Attila analysed the domestic situation and stressed that, before AI can be effectively used in Hungary, a more fundamental issue must be addressed: low levels of digitalisation. Many SMEs still use paper-based records, which creates both efficiency and serious security-policy risks.
He gave the example of employees entering strategic corporate data into public AI tools, which can result in sensitive information leaving the company’s control. Kasznár argued for simultaneous development of digital skills and security awareness, and recommended conscious, state-supported—ideally mandatory—training programmes, since voluntary participation among entrepreneurs is low.
He also noted that corporate security awareness typically focuses on data- and facility-level protections (firewalls, cameras), while cyberthreat levels have risen substantially and society, especially education, is not yet prepared.
Three kinds of AI — three types of risk
Csáki Csaba emphasised that the popular term “AI” covers at least three technologically distinct categories. First are traditional data-driven systems that require organised data assets and domain expertise. Second are generative AI models (language, image, video and music generation), which have a very low user entry threshold and therefore raise different security and regulatory challenges. Third are intelligent agents whose operation demands significant security investments because misconfiguration can quickly lead to runaway costs and risks.
He pointed out that the generative AI market is dominated by a few large players—primarily American and some Chinese or French models—because developing a single new model requires months of work across thousands of servers and billions of dollars in investment. Comparable capital and infrastructure are not available in Europe or Hungary.
Supply-chain dependencies, sovereignty and geopolitics
A central theme of the discussion was the geopolitical entanglement of the entire AI supply chain. Participants noted that more than 60 percent of rare earth mining is under Chinese control, and Chinese interests also oversee significant extraction in South America and Africa. The most advanced chips are produced largely by Taiwanese manufacturers, while key manufacturing equipment technology is controlled by a Dutch firm. Cloud services for AI are dominated by two providers (AWS and Microsoft), and the market for large language models is also concentrated.
This layered dependency network, the speakers argued, goes beyond traditional alliance structures and links to conflicts such as rare-earth aspects of the war in Ukraine. Contemporary wars—particularly the Russia–Ukraine conflict—have accelerated AI development, especially in the use of disinformation as a weapon. The panel agreed that artificially generated content (fake images, false news) now shapes public opinion often regardless of factual accuracy, affecting both warfare and economic communication.
They also discussed quantum computing: the emergence of practical, large-scale quantum machines would fundamentally challenge current public/private-key cryptography and further concentrate technological and economic power in the hands of a few firms and states.
Models of governance and regulation
The discussion compared U.S., Chinese and European approaches. The U.S. model gives large freedom to markets and the private sector; the Chinese model exercises strong state control over data and companies; the European Union seeks an intermediate path, mainly by regulatory instruments (for example GDPR and proposed AI legislation). Participants noted that decisions on trade-offs between security and commercial freedom are often cultural and philosophical as well as technical.
The role of universities: adaptability over narrow credentials
In the final session, speakers addressed higher education’s future role. Csáki argued that rapid, often exponential technological change shifts universities’ responsibility: the emphasis should move away from transmitting narrowly defined specialist knowledge that risks becoming obsolete by graduation, toward developing transferable skills—creative thinking, problem solving, teamwork, communication, self-reflection and adaptability. At the same time, solid foundational domain knowledge remains important to judge whether outputs from AI tools are reliable.
Participants called for better alignment of academic and corporate practical training (for example executive programmes) and tighter integration of universities with economic and social actors.
This summary was prepared on the basis of the roundtable audio and with the help of AI applications.
