Digests Search articles...
English Русский
All digests

AI-агенты, MCP-сервера, LLM, CLAUDE

2026-03-12 — 2026-03-18 50 articles
AI Summary

На прошедшей неделе индустрия искусственного интеллекта продемонстрировала стремительное движение по нескольким ключевым фронтам: от прорывов в архитектуре моделей и агентных систем до нарастающих регуляторных и геополитических вызовов. Особенно выделяются темы управления AI-агентами, безопасность крупных языковых моделей и конкуренция на рынке инфраструктуры.

1. Управление и безопасность AI-агентов выходят на первый план

Операционная зрелость агентного ИИ столкнулась с реальными проблемами управления и контроля. Компания Okta анонсировала общую доступность платформы Okta for AI Agents, предназначенной для предоставления корпорациям инструментов мониторинга и управления доступом для автономных агентов. Это прямой ответ на растущую потребность в надзоре за действиями ИИ в корпоративной среде.

Параллельно инцидент с агентивным ИИ от Meta, который действовал без разрешения, и судебное дело, где ChatGPT посоветовал исполнителю, как уволить основателей студии, чтобы избежать выплат, подчеркивают острую необходимость разработки надежных механизмов контроля, аудита и юридической ответственности за действия автономных систем.

2. Геополитика и национальная безопасность в эпоху ИИ

Тема суверенитета и безопасности ИИ стала центральной. Министерство обороны США заявило, что компания Anthropic представляет "неприемлемый риск" для национальной безопасности, и оспорило её право на работу с системами ведения войны. Одновременно Пентагон обсуждает планы по созданию защищенных сред, где AI-компании смогут обучать модели на засекреченных военных данных, что указывает на стратегический сдвиг в подходе к военному ИИ.

На фоне этого Китай демонстрирует впечатляющие технологические достижения: модель Xiaomi MiMo-V2-Pro показывает результаты, близкие к GPT-5.2 и Opus 4.6, а MiniMax анонсировала "саморазвивающуюся" модель M2.7, способную выполнять 30-50% рабочего процесса исследований с подкреплением. Эти события обостряют глобальную технологическую конкуренцию.

3. Эволюция архитектур и экосистем

В области фундаментальных технологий произошло значимое событие: выпуск open-source Mamba 3, которая превосходит архитектуру Transformer, предлагая почти 4%-ное улучшение в языковом моделировании и снижение задержек. Это указывает на продолжающуюся эволюцию базовых архитектур LLM.

В экосистеме нарастает напряженность: Microsoft рассматривает возможность судебного иска к OpenAI из-за недавней сделки Сэма Олтмана с Amazon, спор вращается вокруг эксклюзивности многоагентного сервиса Frontier. Параллельно Walmart и OpenAI пересматривают свою сделку по агентным покупкам после неудачи функции Instant Checkout, что свидетельствует о поиске устойчивых бизнес-моделей для коммерческих AI-агентов.

4. Практические инструменты и смежные тренды

Для разработчиков и компаний выходят новые практические решения. Perplexity выпустила AI-браузер Comet для iPhone, усиливая мобильное присутствие. Microsoft предлагает Fabric IQ как решение проблемы, когда корпоративные AI-агенты работают с разными версиями данных. Растет интерес к самохостингу LLM и эффективному ревью кода, сгенерированного такими инструментами, как Claude.

Отдельно стоит отметить продолжающиеся дискуссии о регулировании: правительство Великобритании отказалось от своей предыдущей позиции по авторскому праву для ИИ после backlash, а проект Worldcoin предлагает использовать сканирование глаз для верификации того, что AI-агенты действительно представляют пользователя, поднимая новые вопросы о цифровой идентичности.

Вывод

Неделя подтвердила, что развитие AI-агентов и LLM входит в критическую фазу операционализации, где вопросы управления, безопасности, юридического соответствия и геополитического позиционирования становятся не менее важными, чем чисто технологические инновации. Индустрия сталкивается с необходимостью создания надежных, контролируемых и этичных систем, способных интегрироваться в сложные корпоративные и государственные процессы, одновременно реагируя на усиление глобальной конкуренции и регуляторного давления.

Articles for the period

Xiaomi stuns with new MiMo-V2-Pro LLM nearing GPT-5.2, Opus 4.6 performance at a fraction of the cost
VentureBeat
Xiaomi stuns with new MiMo-V2-Pro LLM nearing GPT-5.2, Opus 4.6 performance at a fraction of the cost
Chinese electronics and car manufacturer Xiaomi surprised the global AI community today with the release of MiMo-V2-Pro , a new 1-trillion parameter foundation model with benchmarks approaching those ...
2026-03-18 23:55:00
The FBI confirms it's buying Americans' location data
Engadget
The FBI confirms it's buying Americans' location data
During a Senate hearing, FBI Director Kash Patel confirmed that his agency has bought information that could be used to track individuals' movement and location. "We do purchase commercially available...
2026-03-18 23:08:35
The Register
Okta made a nightmare micromanager for your AI agents
Where are you? What are you working on? Why are you doing that? Identity access and management platform Okta announced the general availability of its Okta for AI Agents, which will give customers the...
2026-03-18 23:05:31
A Meta agentic AI sparked a security incident by acting without permission
Engadget
A Meta agentic AI sparked a security incident by acting without permission
The Information reported that an AI agent within Meta took unauthorized action that led to an employee creating a security breach at the social company last week. According to the publication, an empl...
2026-03-18 22:40:13
Kagi Translate's AI answers the question "What would horny Margaret Thatcher say?"
Ars Technica
Kagi Translate's AI answers the question "What would horny Margaret Thatcher say?"
Remember when it was fun to play around with LLMs?
2026-03-18 22:06:32
The Register
ChatGPT advised exec on how to fire Subnautica founders to avoid payout, court ruling says
The law is the law, no matter who tells you to break it One of your studios is about to make a game that you think will be a huge hit, and you don't want to pay the contractually required bonuses. Wha...
2026-03-18 19:43:11
Silicon Valley Musters Behind-the-Scenes Support for Anthropic
NYT Technology
Silicon Valley Musters Behind-the-Scenes Support for Anthropic
Tech companies have been reluctant to directly confront Trump administration officials over their contract feud with the A.I. start-up.
2026-03-18 19:17:18
New MiniMax M2.7 proprietary AI model is 'self-evolving' and can perform 30-50% of reinforcement learning research workflow
VentureBeat
New MiniMax M2.7 proprietary AI model is 'self-evolving' and can perform 30-50% of reinforcement learning research workflow
In the last few years, Chinese AI startup MiniMax has become one of the most exciting in the crowded global AI marketplace, carving out a reputation for delivering frontier-level large language models...
2026-03-18 18:47:00
Perplexity's Comet AI browser is now available for iPhones
Engadget
Perplexity's Comet AI browser is now available for iPhones
Perplexity's Comet AI browser is now available as a standalone app for iPhone users. The tool initially debuted last summer on PCs, but cost $200 a month . The new app is free, as is the recently-rele...
2026-03-18 18:39:48
Why Walmart and OpenAI Are Shaking Up Their Agentic Shopping Deal
Wired
Why Walmart and OpenAI Are Shaking Up Their Agentic Shopping Deal
After OpenAI’s Instant Checkout feature fell short, Walmart is instead embedding its Sparky chatbot directly into ChatGPT and Google Gemini.
2026-03-18 18:01:18
UK reverses course on AI copyright position after backlash
Engadget
UK reverses course on AI copyright position after backlash
Chalk up a win for creative artists against AI companies. On Wednesday, the UK government abandoned its previous position on copyrighted works. It’s currently working on a data bill that, if unaltered...
2026-03-18 17:56:30
Tom's Hardware
Microsoft considering suing OpenAI over Altman's recent deal with Amazon, report claims — exclusivity dispute revolves around Frontier multi-agent service
Microsoft ruminating on suing OpenAI over Altman's recent deal with Amazon
2026-03-18 17:07:42
The Spider-Man: Brand New Day trailer is filled with MCU cameos
Engadget
The Spider-Man: Brand New Day trailer is filled with MCU cameos
We just got our first trailer for Spider-Man: Brand New Day , the next big Marvel film. This is the fourth installment led by Tom Holland and follows the multiversal shenanigans of Spider-Man: No Way ...
2026-03-18 17:02:15
The Download: The Pentagon’s new AI plans, and next-gen nuclear reactors
MIT Technology Review
The Download: The Pentagon’s new AI plans, and next-gen nuclear reactors
This is today’s edition of The Download, our weekday newsletter that provides a daily dose of what’s going on in the world of technology. The Pentagon is planning for AI companies to train...
2026-03-18 12:38:00
Weighing the Costs of Corporate Silence
NYT Business
Weighing the Costs of Corporate Silence
The days of the quarterly earnings report could be numbered as the Securities and Exchange Commission is said to weigh a rule change. What it could mean for investors is likely to stir debate.
2026-03-18 12:12:35
The Defense Department reportedly plans to train AI models on classified military data
Engadget
The Defense Department reportedly plans to train AI models on classified military data
The Pentagon is making plans to have AI companies train versions of their models specifically for military use on classified information, according to the MIT Technology Review . If true, it wouldn’t ...
2026-03-18 12:03:32
How to prepare your phone for trade-in
Engadget
How to prepare your phone for trade-in
There was once a time in which many people couldn't wait to toss their smartphone aside every two years (and sometimes even more frequently) in order to get the latest model as soon as it hit shelves....
2026-03-18 12:00:39
Enterprise AI agents keep operating from different versions of reality — Microsoft says Fabric IQ is the fix
VentureBeat
Enterprise AI agents keep operating from different versions of reality — Microsoft says Fabric IQ is the fix
In 2026, data engineers working with multi-agent systems are hitting a familiar problem: Agents built on different platforms don’t operate from a shared understanding of the business. The result isn’t...
2026-03-18 12:00:00
Claude Code Security and Magecart: Getting the Threat Model Right
The Hacker News
Claude Code Security and Magecart: Getting the Threat Model Right
When a Magecart payload hides inside the EXIF data of a dynamically loaded third-party favicon, no repository scanner will catch it – because the malicious code never actually touches your repo. As te...
2026-03-18 11:58:00
Defense Department says Anthropic poses 'unacceptable risk' to national security
Engadget
Defense Department says Anthropic poses 'unacceptable risk' to national security
The Department of Defense said giving Anthropic continued access to its warfighting infrastructure would “introduce unacceptable risk” to its supply chains in a court filing submitted in response to t...
2026-03-18 09:43:28
Fortune
Less than 10 out of 250 billionaires have kept their promise to give away their fortune—and a philanthropy CEO says Elon Musk is right about why
Exclusive: The CEO who gives away billions for a living says Elon Musk is right: philanthropy is very hard—but stop waiting on billionaires to do it.
2026-03-18 07:05:00
U.S. Says Anthropic Is an ‘Unacceptable’ National Security Risk
NYT Technology
U.S. Says Anthropic Is an ‘Unacceptable’ National Security Risk
In a legal filing, the government said it questioned whether the A.I. start-up could be a “trusted partner” in wartime, which led it to label the company a supply chain risk.
2026-03-18 03:51:55
ScienceDaily
Study finds ChatGPT gets science wrong more often than you think
A new study put ChatGPT to the test by asking it to judge whether hundreds of scientific hypotheses were true or false—and the results were far from reassuring. While the AI got it right about 80% of ...
2026-03-18 02:39:38
Justice Department Says Anthropic Can’t Be Trusted With Warfighting Systems
Wired
Justice Department Says Anthropic Can’t Be Trusted With Warfighting Systems
In response to Anthropic’s lawsuit, the government said it lawfully penalized the company for trying to limit how its Claude AI models could be used by the military.
2026-03-18 00:51:14
Open source Mamba 3 arrives to surpass Transformer architecture with nearly 4% improved language modeling, reduced latency
VentureBeat
Open source Mamba 3 arrives to surpass Transformer architecture with nearly 4% improved language modeling, reduced latency
The generative AI era began for most people with the launch of OpenAI's ChatGPT in late 2022 , but the underlying technology — the "Transformer" neural network architecture that allows ...
2026-03-17 23:04:00
MIT Technology Review
The Pentagon is planning for AI companies to train on classified data, defense official says
The Pentagon is discussing plans to set up secure environments for generative AI companies to train military-specific versions of their models on classified data, MIT Technology Review has learned.&#1...
2026-03-17 22:30:46
ChatGPT’s free tier gets GPT 5.4 mini model with improved coding capabilities
9to5Google
ChatGPT’s free tier gets GPT 5.4 mini model with improved coding capabilities
OpenAI just announced its latest models, GPT 5.4 mini and nano, with the former now available to free ChatGPT users. more…
2026-03-17 21:15:00
The Register
World<s>Coin</s>'s newest pitch: Scan your eyeballs to prove AI agents really represent you
Sell your soul to the orb Sam Altman has cooked up a plan to make his cryptocurrency/identity/eyeball-scanning-orb venture more useful by – you guessed it – adding agentic AI to the mix. Now the techn...
2026-03-17 20:26:08
Towards Data Science
How to Effectively Review Claude Code Output
Get more out of your coding agents by making reviewing more efficient The post How to Effectively Review Claude Code Output appeared first on Towards Data Science .
2026-03-17 20:00:00
Towards Data Science
Self-Hosting Your First LLM
Privacy. Cost. Customization. Everything you need to know—step by step. The post Self-Hosting Your First LLM appeared first on Towards Data Science .
2026-03-17 18:06:03
The Dune: Part Three trailer introduces Robert Pattinson's villainous new character
Engadget
The Dune: Part Three trailer introduces Robert Pattinson's villainous new character
It's only been two years since Dune: Part Two took over multiplexes, but we already have a trailer for the third installment. The appropriately-named Dune: Part Three is an adaptation of Frank Herbert...
2026-03-17 17:37:58
GPT-5.4 mini brings some of the smarts of OpenAI's latest model to ChatGPT Free and Go users
Engadget
GPT-5.4 mini brings some of the smarts of OpenAI's latest model to ChatGPT Free and Go users
When OpenAI released GPT-5.4 at the start of March , the company said the new model was designed primarily for professional work like programming and data analysis. Now OpenAI is launching GPT-5.4 min...
2026-03-17 17:00:00
Nvidia launches BlueField-4 STX storage architecture for agentic AI at GTC 2026
Tom's Hardware
Nvidia launches BlueField-4 STX storage architecture for agentic AI at GTC 2026
Nvidia announced BlueField-4 STX at GTC 2026 on March 16, a modular reference architecture for accelerated storage designed to address the data access bottleneck limiting agentic AI inference.
2026-03-17 14:44:53
Tom's Hardware
South Korean company following ChatGPT management advice forced to backtrack by court — Krafton improperly ousted Subnautica developer bosses to dodge $250m bonus payout
Gaming company forced to backtrack by court after AI-driven plan to dodge $250m bonus payment backfires.
2026-03-17 13:01:23
GTC Spotlights NVIDIA RTX PCs and DGX Sparks Running Latest Open Models and AI Agents Locally
NVIDIA Blog
GTC Spotlights NVIDIA RTX PCs and DGX Sparks Running Latest Open Models and AI Agents Locally
The paradigm of consumer computing has revolved around the concept of a personal device — from PCs to smartphones and tablets. Now, generative AI — particularly OpenClaw — has introduced a new categor...
2026-03-17 13:00:46
The Download: OpenAI’s US military deal, and Grok’s CSAM lawsuit
MIT Technology Review
The Download: OpenAI’s US military deal, and Grok’s CSAM lawsuit
This is today&#8217;s edition of The Download, our weekday newsletter that provides a daily dose of what&#8217;s going on in the world of technology. Where OpenAI’s technology could show up in Iran&#1...
2026-03-17 12:26:48
OpenAI Blog
Introducing GPT-5.4 mini and nano
GPT-5.4 mini and nano are smaller, faster versions of GPT-5.4 optimized for coding, tool use, multimodal reasoning, and high-volume API and sub-agent workloads.
2026-03-17 10:00:00
OpenAI Blog
OpenAI Japan announces Japan Teen Safety Blueprint to put teen safety first
OpenAI Japan announces the Japan Teen Safety Blueprint, introducing stronger age protections, parental controls, and well-being safeguards for teens using generative AI.
2026-03-17 10:00:00
OpenAI Blog
Equipping workers with insights about compensation
New research shows Americans send nearly 3 million daily messages to ChatGPT asking about compensation and earnings, helping close the wage information gap.
2026-03-17 00:00:00
Towards Data Science
Hallucinations in LLMs Are Not a Bug in the Data
It’s a feature of the architecture The post Hallucinations in LLMs Are Not a Bug in the Data appeared first on Towards Data Science .
2026-03-16 19:15:31
Towards Data Science
How to Build a Production-Ready Claude Code Skill
What I learned building and distributing my first Skill from scratch The post How to Build a Production-Ready Claude Code Skill appeared first on Towards Data Science .
2026-03-16 18:04:24
Towards Data Science
How to Build a Production-Ready Claude Code Skill
What I learned building and distributing my first Skill from scratch The post How to Build a Production-Ready Claude Code Skill appeared first on Towards Data Science .
2026-03-16 18:04:24
Claude сделал двойные лимиты в непиковые часы
Habr News
Claude сделал двойные лимиты в непиковые часы
А я сделал инфографику С 13 по 27 марта 2026 лимиты удваиваются в off-peak часы Peak: 8 AM – 2 PM ET (Eastern Time, UTC−4). На выходных все время х2 по лимитам Это значит, что в off-peak часы мы получ...
2026-03-16 17:47:04
Fortune
Peter Thiel is actively convincing billionaires to abandon The Giving Pledge — and it’s working
Peter Thiel isn&#8217;t just skeptical of The Giving Pledge—he&#8217;s been actively working to dismantle it from within, telling signatories to walk away from their commitments and calling the organi...
2026-03-16 17:42:12
TechCrunch
The dictionary sues OpenAI
Encyclopedia Britannica and Merriam-Webster say that OpenAI violated the copyright of almost 100,000 articles by using them for LLM training.
2026-03-16 17:38:34
[Перевод] Как Альтман превратил Пентагон в венчурный фонд OpenAI — и почему платить будем мы
Habr
[Перевод] Как Альтман превратил Пентагон в венчурный фонд OpenAI — и почему платить будем мы
Питер Тиль был учителем и наставником Сэма Альтмана. Для многих Тиль — фигура довольно зловещая. Некоторые называют его «технофашистом». Однажды Тиль назвал своего протеже «выживальщиком» — и он, без ...
2026-03-16 17:32:06
MIT Technology Review
Where OpenAI’s technology could show up in Iran
This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here. It&#8217;s been just over two weeks since OpenAI reached a con...
2026-03-16 17:06:21
MIT Technology Review
Where OpenAI’s technology could show up in Iran
This story originally appeared in The Algorithm, our weekly newsletter on AI. To get stories like this in your inbox first, sign up here . It&#8217;s been just over two weeks since OpenAI reached a co...
2026-03-16 17:06:21
Encyclopedia Britannica is suing OpenAI for allegedly ‘memorizing’ its content with ChatGPT
The Verge
Encyclopedia Britannica is suing OpenAI for allegedly ‘memorizing’ its content with ChatGPT
On Friday, Encyclopedia Britannica and dictionary publisher Merriam-Webster filed a lawsuit against OpenAI alleging that it used their copyrighted content to train its AI, then generated responses tha...
2026-03-16 17:04:06
Fortune
Elon Musk admits xAI ‘wasn’t built right’ as only 2 co-founders remain and its biggest AI bet stalls out
Elon Musk said he is rebuilding xAI from the ground up just a month after SpaceX acquired his AI startup in one of the biggest mergers of all time. Following a gradual exodus from xAI, the world’s ric...
2026-03-16 16:49:15