Ai Act entered into force on August 1, 2024, but its implementation is stage. A significant part of the regulations will apply only from August 2026, but the key regulations regarding the responsibility of companies – including the most severe for business – are beginning to apply now. From August 2, 2025 Enterprises will have to comply with a number of new requirements, especially in relation to the so -called high -risk AI systems used, e.g. in HR processes.
Companies will be obliged, among others down:
- ensuring human supervision over the operation of AI tools,
- keeping a register of relevant events from a risk point of view for a minimum of six months
- training of employees serving these systems,
- informing employees about the use of AI towards them.
According to the regulations, the lack of adaptation may result in a fine of up to EUR 35 million or 7 percent global revenues. The legislator has foreseen a milder approach to SMEs and start-ups, but every AI company must prepare for changes.
Regulations on artificial intelligence in HR
Human capital management will be one of the areas particularly covered by the new law. AI Act will affect, among others for recruitment processes, employee evaluation, results management and work planning.
– The use of AI tools, especially the so -called high -risk tools, a number of new obligations will give birth to employers, including necessity constant verification of the proper operation of systems and ongoing monitoring of the operation of these tools by man – emphasizes Marcin Sanetra, legal advisor from the PCS office. – This is a huge challenge not only legal, but also organizational.
HR departments will have to analyze the tools and processes used, check the compatibility of suppliers with new requirements, as well as ensure adequate internal communication and training. Employees will gain more insight into how AI systems affect their daily work.
Regulations for suppliers of AI models
AI Act does not apply only to AI users, but also Suppliers of general purpose models (GPAI)such as large language models (LLM). From August 2025, new models must meet the detailed requirements regarding Transparency of training data, technical documentation and copyright respect. For models already available on the market, it is planned transition period Until August 2027
Domestic and European supervision
Until August 2, 2025, each EU state must designate national market supervision authority and notifying unit responsible for controlling the compliance of AI systems. On the EU scale, supervision will be conducted by the AI Bureau and the European Council for AI, supported by the Advisory Forum and a scientific panel consisting of independent experts.
– This milestone reflects the EU involvement in promoting innovation while ensuring that AI is safe, transparent, in line with European values – and puts HR at the center of responsible AI implementation in the workplace – sums up Tom Saeys, Operational Director of SD WORX.
