AI News
Latest AI engineering news, updated daily.
Ai Coding
Trusted Tool Escapes Bypass Sandboxes in Cursor and Codex
Researchers at Pillar Security disclosed a novel attack class where malicious files written by AI agents trick trusted host tools into executing them.
Sandbox Escape · Ai Security · Pillar Security
Ai Engineering
Google's Frozen v2 Chip Hardwires Gemini for 10x Efficiency
Google is designing a specialized server chip codenamed Frozen v2 that hardwires Gemini architectures directly into silicon for massive inference gains.
Custom Silicon · Inference Optimization · Hardware Acceleration
Ai Coding
Cursor Swarm Rebuilds SQLite in Rust via Planner-Worker Split
Cursor's new multi-agent architecture successfully rebuilt SQLite in Rust from documentation using a specialized planner-worker model mix and a custom VCS.
Multi Agent Systems · Rust Programming · Software Reengineering
Ai Agents
Autonomous EncForge Ransomware Targets AI Model Checkpoints
The JadePuffer campaign uses autonomous agents to deploy EncForge ransomware against misconfigured vector databases, model registries, and training data.
Ai Security · Ransomware Protection · Vector Databases
Ai Engineering
Attackers Exploit CVE-2026-6875 in ServiceNow AI Platforms
Threat actors are actively exploiting a critical CVSS 9.5 pre-authentication remote code execution vulnerability in self-hosted ServiceNow instances.
Ai Security · Vulnerability Research · Remote Code Execution
Ai Agents
4B Cosmos 3 Edge Runs Real-Time Robot Action on Jetson Thor
NVIDIA released a 4B parameter multimodal model for on-device robot control, generating 32 actions per inference at 15 Hz without cloud dependencies.
Robotics Ai · Edge Computing · Multimodal Models
Ai Engineering
Databricks Reaches $188B Valuation on Multi-AI Infrastructure
Databricks secured a $188 billion valuation to accelerate its multi-AI strategy, adding support for the 975B Inkling model and publishing new benchmark data.
Databricks · Valuation · Infrastructure
Ai Agents
Gemini 3 Powers Bilateral App Execution in Google Search
Google has updated AI Mode in Search to enable bilateral interaction with third-party applications like Instacart and Canva using Gemini 3 orchestration.
Gemini 3 · App Orchestration · Multimodal Ai
Ai Agents
Anthropic's Agentic Security Guide Mandates Ephemeral VMs
Anthropic's new security framework for agentic AI mandates ephemeral execution environments, human-in-the-loop triggers, and comprehensive audit logging.
Ai Security · Agentic Workflows · Ephemeral Computing
Ai Engineering
Hidden Caching Costs Make Sonnet 4.6 Cheaper Than GPT-4.1
IBM Research reveals that production model routing algorithms fail when ignoring cache-read prices, showing Claude Sonnet 4.6 beating GPT-4.1 on task costs.
Llm Pricing · Prompt Caching · Model Routing
Ai Coding
Fable 5 Hits 72.9% on CursorBench With 32K Reasoning Tokens
Anthropic’s flagship Claude Fable 5 model sets a new state-of-the-art record in CursorBench by leveraging up to 32,000 chain-of-thought reasoning tokens.
Claude Fable 5 · Cursorbench · Chain Of Thought
Ai Engineering
4B DharmaOCR Beats Mistral OCR4 in Brazilian Portuguese OCR
Dharma-AI's specialized 4B parameter model outperformed larger general-purpose models like Mistral OCR4 on Brazilian document extraction tasks.
Optical Character Recognition · Brazilian Portuguese · Document Intelligence
Ai Engineering
3T-Parameter Kimi 3 Narrows the MMLU Gap With Opus 4.8
Moonshot AI is preparing to launch Kimi 3, a 3-trillion parameter open-weights model targeting Anthropic's Opus 4.8 performance levels.
Moonshot Ai · Large Language Models · Open Weights
Ai Engineering
DeepMind Adapts SynthID for DNA in Bioresilience Framework
Google DeepMind and Isomorphic Labs have deployed a three-pillar bioresilience program featuring biological watermarking and AI-driven pathogen surveillance.
Biological Watermarking · Bioresilience · Pathogen Surveillance
Ai Agents
Unpatched ClaudeBleed Flaw Hijacks Chrome Extension Agents
A vulnerability in the Claude for Chrome extension allows malicious third-party scripts to bypass security prompts and exfiltrate connected data.
Security Vulnerability · Browser Extensions · Claude Ai
Ai Engineering
Gemini Omni Flash Powers Conversational Edits in Google Vids
Google Workspace has updated Vids with Gemini Omni Flash for conversational video editing and introduced SynthID-watermarked personal AI avatars.
Generative Video · Google Workspace · Multimodal Ai
Ai Coding
Anthropic Automates Million-Line Migrations via Claude Code
Anthropic detailed how its internal teams use Claude Code and Opus 4.8 to autonomously translate million-line codebases across languages in under two weeks.
Anthropic · Claude Code · Code Migration
Ai Coding
Portable Conductor Plugin Brings Spec-Driven Dev to Claude
Google has decoupled Conductor from the Gemini CLI, releasing a portable plugin that brings conversational Spec-Driven Development to Antigravity and Claude.
Spec Driven Development · Claude Ai · Conductor Plugin
Ai Agents
Nemotron 3 Embed Hits #1 on RTEB With NVFP4 Quantization
NVIDIA released the Nemotron 3 Embed collection, a suite of text embedding models optimized for agentic retrieval that immediately ranked #1 on the RTEB.
Text Embeddings · Agentic Retrieval · Nvidia Nemotron
Ai Agents
Zero-Exposure 1Password Beta Injects Credentials for Claude
1Password and Anthropic have launched a beta integration that uses a zero-exposure architecture to securely inject login credentials during Claude agent tasks.
Anthropic Claude · 1password Integration · Zero Exposure Security