At the third Signatory Taskforce meeting, the EU Commission proposed requiring AI providers to forecast when future systems may surpass current systemic risk tiers using structured methods.
The initiative emphasizes scenario-based testing for harmful manipulation and aggregate industry forecasts to track trends in compute, algorithms, and data, aiming for sector-wide risk visibility.
This approach is part of the EU's broader regulatory strategy to ensure transparency, accountability, and proactive governance, pushing providers to identify and address AI risks before harms occur.
Can the EU's new AI rules adapt quickly enough to govern exponentially advancing technology?
If AI safety rules are delayed, what stops a major AI disaster from happening before 2028?
How can developers prove an AI is safe when its complex reasoning is not fully understood?
As AI systems begin to hire humans, are our laws prepared for the reality of robot bosses?
AI can be secretly poisoned to manipulate you. Is it already too late to trust them?