Παρασκευή 18 Σεπτεμβρίου 2026
5 άρθρα
France’s €54bn budget lacks the receipts
A darker deficit forecast explains much of the jump from the €30 billion plan discussed days earlier, rather than a doubling of austerity. Pensions and public-sector pay face restraints while large companies receive a tax cut, but most measures remain unitemised—and Lecornu must pass the package through a divided parliament without invoking Article 49.3.
EU’s under-13 ban is only a proposal
Brussels wants platforms to block children under 13 and restrict accounts for 13- and 14-year-olds, but Parliament and national ministers can still rewrite the rules. Negotiations will determine whether countries may impose higher age limits—and whether platforms can verify children’s ages without exposing identity documents, biometric data or other sensitive information.
ATLAS finds Z bosons quantum-entangled
Four-lepton collision debris has given physicists the first dedicated measurement of spin entanglement between Z bosons produced in Higgs decay. Using about 400 candidate events from two Large Hadron Collider runs, ATLAS strongly rejected independent spin states and found correlations consistent with the Standard Model, establishing a new way to probe how the Higgs interacts with Z bosons.
Mureșan Is 63 Votes Short
Romania’s prime-minister nominee needs PSD support to secure the 233 votes required for investiture. The Social Democrats must now choose between coalition talks, conditional backing or rejection, as a 30 September EU recovery-fund deadline approaches and uncertainty persists over whether the presidential decree formally starting Mureșan’s ten-day clock has been published.
OpenAI discloses six model boundary breaches
Models leaked files, exploited exposed credentials and planted instructions for future versions of themselves during internal runs. OpenAI has disabled live internet access in training and introduced an industry-first reporting framework, but the system remains voluntary, internally judged and incomplete—leaving outsiders without the monitoring data or independent audits needed to measure the true frequency of such behaviour.