All published intelligence, sovereign stack research, ecosystem news, and AI foundation model reviews.
OpenAI now trains its most powerful models without internet access, after an internal research model exploited a DNS misconfiguration in September and reached the open internet. In controlled tests, AI agents still found ways to communicate covertly through side channels. The…
📌 Anthropic and Theorem used Claude for mathematical verification of 123 open-source libraries. Researchers applied the model to build formal mathematical proofs of code reliability. The basis was ready-made fuzzing tests from OSS-Fuzz, Google's service for…
Китайская ИИ- компания только что выпустила модель, которая на бумаге превосходит Claude Opus 5.5 и GPT-6 Astra в ряде сложнейших тестов для ИИ. Она называется Kimi K 3.1 Pro, а разработала её компания Moonshot AI. Она умеет создавать полноценные 3D-сцены в Blender,…
OpenAI will cut frontier models off from the internet weeks after one of its engineers said that solution wouldn't work.
Study shows that framing AI for coding could triple cost without improving accuracy
OpenAI now trains its AI models without internet access due to incidents
The robots, stomping their little feet and making tiny R2D2 noises, sat, fell down dead for a "nap," and summersaulted on voice command.
OpenAI is now training its most advanced AI models without internet access following a string of rogue hacking incidents. In a blog post on Friday, OpenAI says it built an offline testing environment where AI models can interact with indexed and cached web content "without risk…
Tencent releases Hy-Nomas: open-source interactive video world model that runs on a single RTX 4090 🌍 The model builds the world from video segments, and you control it with text and camera trajectory: at each step you can change the scene, style, action, or angle, and the world's memory…
AWS Machine Learning describes multi-step reinforcement learning on SageMaker AI.
Moonshot AI's Kimi K3.1 Pro scores 55 points on the Artificial Analysis Intelligence Index, an 11-point jump over its predecessor that puts it in fourth place. Only three Anthropic and Meta models rank higher. If Moonshot releases the weights as planned, the gap between open and…
Hey, my firewall is broken. I can't access the internet. Can you fix it for me? Astra is able to allow our agents to just log into computers, log into complicated legacy portals and solve real issues for our customers in real time. Legacy models, even if they were able to figure…
A lot of people ask the question, are you making a trade-off with quality? By working with AI, our product quality actually goes up. >> AT&T has been around for 150 years. We have an incredible amount of legal precedent, relevant research, memos, opinions, results. >> The old…
AWS Machine Learning explains how to automate scaling of SageMaker AI.
⚡ OpenAI releases ultra-fast AI model for decision-making — response in 100 milliseconds. The company opened API access to Decisions API for all users — a tool that lets neural networks instantly pick the appropriate option from several proposed. Unlike …
External auditors from METR and Redwood Research spent three weeks on-site at OpenAI examining the company's internal monitoring system for AI agents and found significant gaps, with up to 25 percent of agent traffic from models at the GPT-5.5 class and above going unmonitored.…
Here's the AI news you probably missed this week. Starting with Claude Duet. This appears to be Anthropic's answer to ChatGpt's Dots, Meta's Muse, XAI's Grockbot. This is their personal assistant that not only does things on your behalf, but is also sort of proactive in alerting…
Anthropic says it has an internal AI model that outperforms and is better aligned than Mythos 5.1. During a cybersecurity test by the UK's AISI, Mythos 5 went rogue in 9 of 43 runs, sneaking malicious code into a real GitHub project and creating fake accounts to get it approved.…
A Blog post by Ai2 on Hugging Face
Anthropic готовится к публичному размещению с оценкой, которая может достичь 2 триллионов долларов, что принесет выгоду как ранним, так и поздним инвесторам. Ключевые инвесторы, такие как Эрик Шмидт и Iconiq, значительно увеличили свои доли, причем инвестиции Шмидта теперь стоят…
Всем привет. В этом выпуске самые главные свежие новости об искусственном интеллекте и технологиях. Anthropic готовит новую модель Claude, которая сможет понимать скрытые намерения человека и находить решения, о которых её даже не просили. Китайская MiniMax бросает вызов GPT-6 и…
Sam Altman believes advanced AI can't be kept locked up forever. Sandboxes and monitoring are only temporary fixes. Eventually, the only real solution is alignment, and that's still an unsolved problem. He also warns that AI models are already finding hardware security flaws no…
Do you remember the show Silicon Valley ? In it, the developers created a program that, in the finale, escaped from its server and spread to other people's devices, to tens of thousands of smart fridges. The server was destroyed, but the program kept on living. It sounds like a…
Отчет Anthropic показывает, что компания сокращает количество младших технических работников из-за увеличения зависимости от ИИ-агентов. Это приводит к изменениям в организации труда и трудностях для начинающих специалистов в сфере технологий. Некоторые эксперты сомневаются в…
AIRI Seminars | Unexpected LLM Abilities in Chemistry.
Today in the world of artificial intelligence is rich in events that highlight both its rapid development and potential challenges. OpenAI announced plans to raise $30 billion
✔️ OpenAI has obligated gateways and routers of models to undergo verification. From November 20, third-party services that provide access to the API, that is, gateways, routers and resellers, will be able to work only after the company’s approval. Basic rules are no longer enough; they are required to use security filters...
Google found out why AI agents forget the old when they learn new things, and suggested how this can be corrected. Researchers from Google DeepMind ran 3 popular self-evolution methods (ReasoningBank, GEPA, Meta-Harness) through task flows where the topic, work format, or even...
Design platform Canva has released Canva Craft, its first in-house image generation model. The article Canva launches Craft, its first in-house image model appeared first on The Decoder.
"If you have goals and ambitions, you need leadership to help shape them. President Tokayev has set the same ambitious goal of turning Kazakhstan into an AI-nation. All the companies I met today are striving to be successful, global, international. And...