This week there was a significant development in the progression of the AI ACT. According to a report, the EU Council is close to finishing their compromise text and could reach an agreement next month.
Once finalized, the text will form the basis of the Council’s position in the trilogy negotiations between the EU’s three institutions, which are expected to start in the early part of next year.
Scope
- Biometric recognition
- General-purpose AI
- Law enforcement
- Innovation
- Governance
- Transparency obligations
- High-risk AI systems
A new transparency obligation has been added, requesting the providers of systems susceptible of causing significant harm to include the expected output in the instructions for use when appropriate.
Penalties
The list of violations that could lead to fines of up to 20 million or 4% of annual turnover has been expanded to include the violation of transparency obligations.
When calculating the sanction, more points were added for the intentional or negligent nature of the violation, attempts at mitigation, and if similar violations have already been sanctioned.
Common specifications
The approach followed on common specifications was aligned with the Regulation on machinery. In particular, the common standards would be repealed once harmonized standards were adopted.
Read the whole article here