European Parliament approves the AI Act in plenary vote

On 13 March 2024 the European Parliament voted through the Artificial Intelligence Act, the text agreed with member states the previous December. The count was 523 in favour, 46 against and 49 abstentions.

What it is

The plenary vote was Parliament's formal approval, but two steps remained. Parliament said the text still faced a final lawyer-linguist check and needed formal endorsement by the Council before it could be published.

What it requires

Parliament's release set out the phase-in that security and compliance teams still plan around. Bans on prohibited practices would bite six months after entry into force, general-purpose AI rules and governance after 12 months, and high-risk obligations after 36 months. High-risk systems must assess and reduce risks, keep use logs, be transparent and accurate, and allow human oversight. The more powerful general-purpose models face extra duties: model evaluations, systemic-risk mitigation and incident reporting.

What security leaders should do

Use-log retention is the requirement most likely to land on a security team's desk. Check whether your logging pipeline can capture inputs, outputs and operator actions for any AI system that might be classed as high-risk, and whether retention settings would survive an audit request. Pair that with a named owner for human oversight, so the log has someone accountable for reviewing it. Note that later amendments moved some high-risk dates, so confirm current deadlines against the Commission's timeline before committing budgets.