The application of the European Union’s artificial intelligence regulation, the AI Act, entered a new phase in August. A panel at the HVG AI Talks conference heard that the requirements include making it clear to users when they are communicating with AI through a chatbot. Companies must also ensure that employees who use AI are adequately prepared.
Application of the rules for high-risk systems, however, has been postponed until the end of 2027 and 2028, respectively. Panel participants said regulation should not restrict the technology for its own sake, but establish a framework for its safe use.
Early adoption and data security risks
Horváth-Brunner Rita, marketing and communications director at VOSZ, said 116 small and medium-sized businesses had been approached across Hungary. While some had already integrated AI deeply into their operations, 64% were still taking their first steps, she said.
Many managers do not know whether their employees use AI tools at work. This can create data security risks if staff upload business information to external services through personal accounts. Horváth-Brunner said businesses should therefore set clear rules on the purposes for which the technology may be used and the data that may be entered.
The human side of adoption
Opten began introducing artificial intelligence gradually three years ago. According to strategy director Csorbai Hajnalka, the company uses the technology both to support internal operations and in services offered to clients.
The company has launched more than 300 pilot projects in recent years. Its experience suggests that adoption is now at least as much a human resources management issue as an IT issue: initial employee resistance needs to be turned into curiosity, while concerns about whether AI could threaten jobs over the longer term must also be addressed.
Legal compliance does not carry over automatically
Badinszky Áron, the rector’s commissioner for artificial intelligence at Eötvös Loránd University (ELTE), warned that using an AI tool does not automatically exempt a business from legal obligations. A large language model developer’s compliance with the AI Act does not necessarily mean an application built on that model is compliant; a company using the finished system may also have obligations.
In Hungary, compliance is monitored by the AI Market Surveillance Authority and the relevant sectoral regulators. The authorities already have the necessary powers, but there is still limited experience of how those powers work in practice.
At the conference, Marco Mongiello, head of the business school at London’s Regent’s University, also cautioned that although many companies had invested substantial sums in AI, those investments had not necessarily produced tangible business results. Participants said that choosing tools was therefore only part of the task: companies must also prepare employees and prevent efforts to improve efficiency from creating new risks.



