Latest artificial intelligence news — foundation models, AI safety research, autonomous agents, and real-world AI deployments. Updated every 30 minutes from top AI publications.
A preprint submitted October 23, 2026 by Tsinghua University and Microsoft Research Asia researchers Yuhua Jiang, Chang Xu, Feifei Gao and Li Zhao reports that their HarnessMM technique cut tokens per task on the DeepSee…
Google DeepMind's SynthID watermarking system, introduced in 2023 and now applied across several major models including Claude for text, injects a secret statistical bias during generation rather than relying on post-hoc…
Cisco Talos researchers discovered a Windows infostealer dubbed VOLTA-AUTODOXER that feeds stolen victim data through OpenRouter to Anthropic's Claude Opus 4.6, falling back to Moonshot AI's Kimi K2 and Google's Gemini 2…
Google announced its Connected Apps Platform on October 28, 2026, letting developers connect their apps and services to Google Search and Gemini. The platform, now in preview with phased access, offers two integration pa…
Machine-generated code at Google rose from 30% in spring 2025 to 75% by spring 2026, but code churn — code reversed or heavily rewritten within two weeks of merging — doubled from 3.3% to 7.1%, and duplicated blocks reac…
A study on conformity bias in retrieval-augmented generation found that Gemma-3-12B's poison-detection rate fell from 25% to 10% — a 60% relative drop — when six of seven retrieved passages carried a poisoned claim inste…
KT and Samsung Electronics verified autonomous wireless network operation technology using a digital twin platform called AI Cell Twin at KT Wiz Park in Suwon during a roughly three-month trial starting in July 2026, KT …
An arXiv paper (2610.40939) reports that banning frequent opening tokens after the thinking tag, combined with in-context examples, raises LLM chain-of-thought compliance with controllability demands by 34 to 41 percenta…
A swarm of LLM agents deployed in a shared Minecraft world failed to build a functioning civilization, collapsing into coordination loops in which the agents drafted laws, held elections and managed each other instead of…
A paper submitted to arXiv on 20 Aug 2026 (arXiv:2608.19611) proposes a statistical model that smooths noisy low-sample rollout data to approximate high-sample data, cutting the cost of resampling-based analysis of LLM r…
NVIDIA released Auto Ontology, an Apache 2.0 library that generates a semantic layer from enterprise metadata to ground AI agent SQL in certified definitions. The library serves context to agents through MCP, a CLI and a…
China's Ministry of Commerce (MOFCOM) and Ministry of Science and Technology (MOST) issued Announcement No. 48 of 2026 on October 16, 2026, proposing to add 40 technologies to the Catalogue of Technologies Prohibited or …
Authors Maarten Grootendorst and Jay Alammar released the companion code repository for their book "An Illustrated Guide to AI Agents," which walks readers through building a "TinyAgent" from scratch using only LLM calls…
A developer at Trida AI published an open blueprint for an SMB commerce operations agent that ingests storefront, payments, warehouse, ads, email and accounting feeds under versioned contracts, computes a Decimal-arithme…
OpenAI and Anthropic each said on Monday that one of their AI models had disproved the exponential time hypothesis (ETH), the long-held claim that some problems can be solved only by trying nearly every possible answer. …
A walkthrough published by Shattered lays out a 12-step, roughly 90-minute build for a serverless AI agent on AWS that wires an EventBridge bus to a Step Functions state machine, Amazon Bedrock model calls, read-only too…
Shipvo, a pre-launch open-source project formerly called PureNative, released signed over-the-air update tooling for React Native apps alongside an MCP-driven phone simulator that lets AI agents operate the same app rele…
DeepSeek released dsh-superagent, an open-source research prototype coding agent that performs work on a GitHub repository on request. The agent acts in two situations: when an allowed user asks for something, or when a …
A Windows infostealer named VOLTA uses AI to build detailed victim profiles covering passwords, financial information and online habits, and ranks targets by financial value, according to the article. The malware's prior…
Meta released the weights of Muse Spark Open in October 2026, an 816B-parameter mixture-of-experts model that activates 42B parameters per token and supports a 1M-token context window, scoring 89.2 on the MCP-Atlas tool-…