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
The Dutch offensive security firm has raised its total funding to $65 million. Its AI agents map the attack surface and verify which vulnerabilities are actually exploitable.
Led by Oak HC/FT, Vitalize’s Series A funding round is supporting a platform that calculates the most cost-effective allocation of care staff in real time; the platform is already in use in 37 hospitals.
Speaking before an Australian parliamentary committee, Jason Kwon acknowledged that OpenAI should have notified Canberra sooner. Anthropic, for its part, supports a requirement to report serious incidents.
Seoul is splitting its sovereign model programme into two parts and launching a new competition open to start-ups. The budget, amounting to around 3.5 billion dollars, has yet to be approved.
Google’s largest electricity contract on the PJM network combines an increase in capacity equivalent to 11 nuclear reactors with a long-term supply agreement, in response to the regulations imposed on data centres.
Valued at one billion from the outset, the company aims to have data, voice recordings and images ‘signed off’ by their creators, and to pay them each time they are used, based on a specific model.
Rare lesions – accounting for less than 1 per cent of images – almost always escape detection by the models. By taking the rest of the video into account, the HCE method increases their recall by a factor of 2.7 to 3.1.
Steve Croft describes how machine learning enables Breakthrough Listen to search through archives and real-time data streams – including those from the Vera Rubin Observatory – in the hunt for technosignatures.
▶
Charles Frye (Modal) suit une requête à travers un moteur d'inférence comme SGLang ou vLLM, et détaille cache KV, CUDA graphs, décodage spéculatif et débogage en production.
▶
Jim Clark (Docker) explique pourquoi la sécurité d'un agent tient à son bac à sable et aux outils qu'on y fait entrer, et pourquoi il ne devrait y détenir aucun identifiant.
Using the ToyTalk tool, pupils set up a chatbot for their toy and then test it: they want caring companions, but rarely reconfigure the chatbot when things go wrong.
Rather than carrying a load, the robot tows it: a single control system manages starting, turning and stopping, whether with human passengers or not, and without the need for any specific adjustments.
Combining an ultrasound image classifier with another classifier trained on simple clinical data increases the AUROC from 0.91 to 0.95, without the need for costly tests.
The protocol treats physical, hybrid and purely data-driven models equally, testing them against ten years of observations from after 2015 and against past climates.
More than 15 New Yorker cartoonists have recognised their own signatures on images generated by ChatGPT: beyond stylistic imitation, this raises the question of attribution.
The new version of Cohere's agent platform redesigns the orchestration, adds persistent memory, and allows administrators to cap consumption per user.
After intrusions into Shinhan, KB Kookmin, Hana and Woori, Lee Jae Myung evokes "signs" of AI use and calls for cybersecurity adapted to this new era.
To comply with the transparency rules set out in the AI Act, OpenAI is rolling out textGrain, an invisible watermark embedded in the choice of words, exclusively for European users.
The Chinese laboratory is reportedly targeting 80 billion yuan, well above the 50 billion initially hoped, in preparation for an IPO.
▶
Un ingénieur de PayPal compare trois façons d'encaisser un paiement lancé par un agent IA : Instant Checkout dans ChatGPT, AI Mode de Google et une application MCP avec page de paiement externe.