AM-IA brings together news, research papers and applications relating to artificial intelligence published around the world into a single feed — and makes them available in your European language at the click of a button.
Scroll to browse · Click to filter
Its *World Development Report 2026* estimates that AI could enable developing countries to achieve in ten years what would normally take a century.
Faced with the rise in electricity consumption linked to AI, researchers and industry are turning to AI-optimised carbon capture materials, but real-world demonstrations are still limited to small-scale trials.
OpenAI has announced that it has surpassed the one billion mark for active users, whilst significantly reducing the prices for its GPT-5.6 Luna model, driven by efficiency gains achieved through the AI itself.
Cloudflare Wallets and cloudflare.pay provide AI agents with a stable identity and a stablecoin wallet governed by spending limits set by their creators.
Since 2 August 2026, the European AI Office has been able to request technical documentation for models, assess their risks and impose fines of up to 15 million euros or 3 per cent of global turnover.
The time saved thanks to generative AI does not translate into more rest for employees, but rather into higher production targets — and a new, hidden workload of monitoring staff.
▶
Alan D. Thompson passe en revue en direct la mise à jour de son tableau comparatif des grands modèles, référence suivie pour tracer l'évolution des tailles, des dates et des capacités.
Unit 42 was able to observe from inside an offensive campaign led by DeepSeek via the Hermes Agent frame: the agent had exposed his own file server.
The randomized SAVE-O2 AI trial, conducted in 300 patients in four U.S. hospitals, shows that an automated oxygen flow adjustment holds the therapeutic range much better than caregivers.
The benchmark evaluates the customization of agents from long interaction histories rather than frozen profiles, and reveals a "gaps between knowledge and action".
Marta Garnelo and Wojciech Czarnecki rule out four common explanations and show that a leading model is the only one, on nine methods, whose accuracy drops when the number of dimensions increases.
The F-ICL test bench compares the induction in the context of the models to an exact computable standard: the accuracy of the responses masks an inductive bias that the scale does not correct.
The benchmark covers three types of graphs, four task families and 84 executable tools. Authors' conclusion: the quality of tool calls counts more than their number.
Passed by a vote of 82 to 16 in a mock student senate, the "STUDENTS FIRST Act of 2026" rejects both a total ban and unlimited adoption.
▶
Alors qu'Anthropic, OpenAI et Moonshot enchaînent les sorties de modèles phares, cette vidéo interroge le silence relatif de Google sur sa prochaine génération de Gemini.
The largest model ever produced by Alibaba is based on a mix of experts and only activates 95 billion parameters per token generated.
▶
Premier épisode d'une nouvelle série hebdomadaire, consacré aux preuves mathématiques produites par le modèle Astra d'OpenAI et à la sortie de Gemini Robotics 2.
▶
Session plénière de l'après-midi du Agentic AI Summit 2026 organisé par Berkeley RDI, réunissant chercheurs et dirigeants autour des agents IA.
Age verification by conversational AI operators, $25,000 in statutory damages by digital imitation, mandatory clinical examination in care decisions.
Since 2 August, the European Commission has officially applied the strictest rules of its regulation on artificial intelligence: transparency, tagging of content and high-risk systems.