AI use
The rules for all use of AI in the company: inventory, policy and course. And where you stand if the supervisory authority or an auditor asks. How to
2 of 6
Ready for the regulator
4 items missing
Missing (3) · not urgent today
- Adopt the AI policy (draft, version 1)GoalCube has written it. Read it and adopt it.
- Take the AI literacy course (10 min.)Everyone who uses AI takes the 10-minute course and gets a proof of completion.
- Review the risk for the people concerned (draft ready)A short analysis of the risk to the people the data is about. GoalCube has made the draft; you edit and save it.
Mark as done
Where you stand
- All use of AI is recorded and assessedMissing: 1 waiting for confirmation, 1 task from the legal requirements.See the tasks
- The AI policy has been adoptedNo AI policy adopted yet.
- Those who use AI have taken the courseTake the AI literacy course (10 min.)
- A person decides cases about people1 case waiting for a person.Decide
- Everything is saved as proof12 events in the audit log. None has been changed or deleted.
- The AI has only seen codes, not your dataReal names and numbers have stayed with GoalCube in the EU.
Background, deadlines and shortcuts
Also here: Templates · AI providers · Report for the regulator
Ready for 2 December 2027 (3)
Requirements for high-risk AI, e.g. in recruitment. They apply from 2 December 2027, but you can start on them now.
- Assessment of applicants for controller: Use the system according to the provider's instructions for use and monitor its operation. (Art. 26(1) and (5) AI Act)
- Assessment of applicants for controller: Tell the people the decisions concern that AI is involved. (Art. 26(11) AI Act) · Use the template
- Assessment of applicants for controller: Inform employee representatives and the employees affected before the system is put into use. (Art. 26(7) AI Act) · Use the template
What GoalCube does for you (4)
- Assessment of applicants for controller: Human oversight by a competent person with authority, who genuinely assesses the AI's result. GoalCube People decisions: a person decides each case
- Assessment of applicants for controller: Keep logs for at least 6 months. GoalCube's tamper-evident log
- Assessment of applicants for controller: Explain the decision if the person asks for it. The reason is saved with each decision
- Assessment of applicants for controller: Decisions with a significant effect on people must not be made solely by automated means. The person has the right to human intervention and to contest the decision. GoalCube People decisions: a person decides each case and gives a reason
Deadlines
- 25/5/2018GDPR applies, including Art. 22 on automated decisions and Art. 35 on DPIAs. (GDPR), applies now
- 2/2/2025The rules on prohibited AI practices and AI literacy apply. (Art. 4 and 5 AI Act), applies now
- 2/8/2026Transparency duties for chatbots and AI-generated content apply. (Art. 50 AI Act), applies now
- 2/12/2027Requirements for the use of high-risk AI (e.g. recruitment and HR cases) apply. (Art. 26 and Annex III AI Act (Digital Omnibus)), coming
What is at stake
The maximum fines under the AI Act and GDPR. For small and medium-sized businesses, the lower of the two amounts in the AI Act applies. The fine depends on how serious the infringement is and what the company has done to avoid it. What you do in GoalCube shows your efforts.
- €35 million or 7% of global turnoverProhibited use of AI (Art. 99(3) AI Act)
- €15 million or 3%Other AI Act requirements, e.g. human oversight, logs and transparency (Art. 99(4) AI Act)
- €20 million or 4%Automated decisions, data subjects' rights and transfers outside the EU (Art. 83(5) GDPR)
- €10 million or 2%E.g. a missing DPIA or data protection by design (Art. 83(4) GDPR)
GoalCube rule set 2026.3
When the rules change, the entries that need to be confirmed again are added to the task list.
- 9/10/2026Deadlines: high-risk requirements from 2 December 2027 under the Digital Omnibus; the transparency duty (Art. 50) applies from 2 August 2026.
- 8/10/2026Art. 22 GDPR requires a human decision, even outside Annex III of the AI Act. A DPIA is required for decisions about people and for profiling.
- 7/10/2026First set of rules: prohibited use (Art. 5), high-risk (Art. 6 and Annex III), transparency duty (Art. 50) and AI literacy (Art. 4).
Simplified and not legal advice.