Key high-risk requirements for employment AI apply from 2 December 2027
Following the 2026 amendments, 2 December 2027 is an important date for Annex III high-risk systems.
Knowledge Hub
Knowledge becomes useful when it can be found, understood, verified and applied.
The Knowledge Hub brings together research, policy proposals, methodologies and learning from pilots. Every item states its audience, evidence status, version, authorship, funding and recommended use.
Browse by question
How is AI changing work? How do we design a fair pilot? How can job quality be measured? How should institutions respond to questions? What supports professional transition?
Use policy
Materials may be used under the stated licence and attribution requirements. Methods include limitations and context; they should not be treated as universal recipes.
Knowledge becomes useful when it can be found, understood, verified and applied.
AI for recruitment, promotion, termination and evaluation is a high-risk category
The AI Act places a range of employment and worker-management systems among high-risk use cases.
Bulgaria’s government programme sets August 2027 for the AI sandbox
The national programme provides for an AI regulatory sandbox with an August 2027 deadline.
The deadline for a national AI regulatory sandbox is 2 August 2027
The current AI Act requires at least one national sandbox to become operational by 2 August 2027.
A national AI Act implementation law is planned alongside the strategy
The government states that a new national law implementing the AI Act is planned for autumn 2026.
Bulgaria’s national AI strategy is planned for autumn 2026
The government has announced national Apply AI and AI First policies.
28.0% point to the deploying organisation as responsible for an AI error
Responsibility is not concentrated in one actor: 28.0% point to the deploying organisation, 27.7% to system creators and 15.8% to the human user.
13.2% support a duty to inform and consult workers when introducing AI
Worker participation is recognisable as a distinct policy option.
13.9% support shorter working time without lower pay when AI takes over tasks
The result provides Bulgarian empirical evidence relevant to how productivity gains might be shared.
16.5% want stronger state oversight of safe and ethical workplace AI
Some respondents see public oversight as a necessary element of protection.
Data & methods
The data & method repository will be filled once real materials pass editorial review and receive a licence. We do not publish materials before verification.