#искусственный-интеллект
-
68% of Peer-Reviewed Papers Turned Out to Be AI Slop: A View from the Trenches
Two reviewers for NeurIPS, WACV and an ECCV workshop went through 22 submitted papers and uncovered fabricated citations and impostor authors on a massive scale. Two papers with fake names were still accepted for oral presentations.
-
AI Helped Chrome Close 1,072 Vulnerabilities Across Two Releases
Google claims that language models have dramatically sped up the discovery and patching of holes in the browser: versions 149 and 150 fixed more bugs than the previous 23 releases combined.
-
Gemini Robotics 2 controls a humanoid's body from feet to fingertips
Google DeepMind has unveiled an updated model, Gemini Robotics 2, which now coordinates the movements of an entire humanoid robot rather than just its upper half.
-
The LLM2HUMAN Clinic: Turning Language Models Into Living People for $19.95
A satirical site in the aesthetic of GeoCities promises neural networks arms, legs and the legal right to order a sandwich — while slyly poking fun at the whole AI industry.
-
Nadella sets a course to compete with OpenAI and Anthropic
Against the backdrop of record earnings at Microsoft, the company's chief openly pitches businesses its own models and agents as an alternative to the very labs it has invested in.
-
GCC Will Reject Any Legally Significant AI-Generated Contribution — Except Tests
The GCC steering committee has approved a policy barring “legally significant” code generated by language models from the compiler. The only exception is for test cases.
-
Claude Independently Derived an Attack on the Post-Quantum HAWK Signature and Sped Up the Cracking of 7-Round AES
Anthropic reports that its Claude Mythos Preview model almost single-handedly found a key-recovery attack on HAWK-256 and accelerated a known attack on AES-128 by a factor of 200–800. The results, however, do not affect production systems.
-
The Largest US Power Grid Will Be Saved from Blackouts by Switching Off Data Centers
Grid operator PJM Interconnection has said that when capacity runs short it will begin temporarily cutting power to large data centers. The reason is the avalanche of new construction, which generation cannot keep pace with.
-
STAR Labs Turned a Race Condition in Linux's Traffic-Control Subsystem Into a Root Exploit With the Help of AI
Researcher Li Jia Jie says artificial intelligence helped him find a use-after-free vulnerability (CVE-2026-53264) and speed up the development of an exploit that grants a local user root privileges.
-
Court Dismisses Google's DMCA Suit Against Web Scraper SerpApi
Google and Reddit are trying to use copyright law to block the scraping of search results, but the court offered a reminder: search results aren't protected by copyright, and the platforms themselves don't own the internet.
-
NVIDIA and Two Dozen Companies Form the Open Secure AI Alliance
An industry alliance led by NVIDIA aims to make open AI models safer and give defenders tools they can inspect and run for themselves.
-
One Downed Power Line Exposed the Grid's Problem in the Age of AI Data Centers
A snapped transmission line outside Washington nearly triggered a failure in the largest power grid in the United States: more than three gigawatts of data centers dropped off the grid all at once, sending voltage swinging from Virginia to Chicago.
-
OpenAI Confirms Global ChatGPT Outage
On the morning of July 25, ChatGPT stopped loading conversations worldwide. The outage also hit OpenAI's API and the Codex coding platform.
-
OpenAI Micro: the $230 AI Keypad Coders Aren't Rushing to Buy
OpenAI's first piece of hardware is the Micro, a compact macropad add-on for working with ChatGPT and Codex. A TechCrunch reviewer found it fun but never quite figured out who actually needs it.
-
Midjourney Buys the Astrology App Co-Star
Midjourney, the AI startup previously known for generating pictures, has acquired the popular horoscope app Co-Star. The service's founder has also become Midjourney's head of design.
-
Anthropic Releases Opus 5 — Nearly on Par with Fable 5, but Half the Price
Anthropic has unveiled Claude Opus 5: the company promises capabilities close to its flagship Fable 5, stronger protection against cyberthreats, and the same price. The release comes amid tightening AI regulation.
-
Black Forest Labs Unveils FLUX 3 — a Multimodal Model for Images, Video and Sound
Germany's Black Forest Labs has opened early access to FLUX 3 — a single model that learns from images, video and audio at once in order to build a picture of the physical world.
-
Kimi K3 AI agents found flaws in Redis and assembled a working RCE exploit
On July 23, Redis shipped seven fixes after the publication of PoCs demonstrating remote code execution against stock builds 6.2.22, 7.4.9, 8.6.4 and 8.8.0. The researchers claim the chains were found by autonomous AI agents.
-
Dolphin X Trojan Uses AI to Rank Victims by Value
A new infostealer with an “AI Profiler” feature automatically scores infected machines, helping attackers pick the most lucrative targets. The feature was spotted by researchers at Varonis Threat Labs.
-
AMD Unveils Instinct MI455X Accelerator and Helios Rack for AI
At the Advancing AI 2026 conference, Lisa Su announced the Instinct MI455X accelerator with 432 GB of HBM4 memory and the production-ready Helios platform.
- More