AI News
Latest AI engineering news, updated daily.
Ai Agents
Agent Search Queries Leak Private Data in MosaicLeaks Study
Researchers from ServiceNow AI and the University of Edinburgh found that deep-research AI agents leak sensitive internal data through external search logs.
Data Privacy · Security Vulnerability · Deep Research Agents
Ai Coding
Claude Code v2.1.181 Renders Terminal Sessions as Live Web Pages
Anthropic's terminal agent now transforms command-line workflows and MCP tool outputs into interactive, shareable web pages hosted on the Claude platform.
Anthropic · Claude Code · Terminal Agent
Ai Engineering
3nm Trainium3 Chips Pivot AWS to Direct Merchant Silicon
Amazon Web Services is shifting its semiconductor strategy, selling its 3nm Trainium3 and Inferentia3 AI chips directly to external data center operators.
Aws Trainium · Semiconductor Industry · Cloud Infrastructure
Ai Agents
DeepMind Cuts Agent Jailbreaks 94% via Supervisor Architecture
DeepMind’s new AI Control Roadmap shifts agent safety from passive guardrails to real-time sandboxing, reducing jailbreak success rates by 94%.
Ai Safety · Jailbreak Prevention · Agentic Systems
Ai Engineering
500,000 Sensors Power Midjourney's Petaflop Ultrasonic Scanner
Midjourney Medical has unveiled a full-body ultrasonic CT scanner that processes two petaflops of data to map internal tissues in under 60 seconds.
Ultrasonic Scanning · Petaflop Computing · Medical Imaging
Ai Agents
Ask Pinterest App Tests Agentic Visual Search in US Sandbox
Pinterest launched Ask Pinterest, an experimental AI shopping app, alongside new enterprise MCP tools following a $4 billion AWS infrastructure deal.
Visual Search · Shopping Assistant · Agentic Ai
Ai Engineering
Meta AI Mode Grounds Search in Social Data via Llama 4
Meta's new AI Mode uses a fine-tuned Llama 4 model and RAG pipeline to synthesize public Facebook and Instagram posts into generative search responses.
Meta Ai · Llama 4 · Retrieval Augmented Generation
Ai Agents
$99 Google Home Speaker Retires Assistant for Gemini
Google launched the $99.99 Google Home Speaker, featuring a dedicated NPU to shift the ecosystem from legacy Assistant to the generative Gemini for Home.
Generative Ai · Smart Home · Gemini Ai
Ai Engineering
XDOF Exits Stealth With $70M and 130K-Trajectory Robot Dataset
XDOF raised $70 million to build a three-tier physical data collection pipeline and co-released the massive ABC-130K manipulation dataset with UC Berkeley.
Robotics · Physical Ai · Dataset Release
Ai Engineering
Pramaana's $27M Seed Brings LEAN Formal Verification to LLMs
Pramaana Labs secured a $27 million seed round to build a deterministic verification layer that uses the Lean programming language to prove AI outputs.
Formal Verification · Lean Programming · Deterministic Ai
Ai Agents
Flue Integration Opens Cloudflare Agents SDK to Custom Harnesses
Cloudflare has decoupled its Agents SDK into foundational platform primitives, allowing external frameworks like Flue to orchestrate stateful AI workflows.
Cloudflare Agents Sdk · Ai Orchestration · Stateful Workflows
Ai Agents
Open ARD Specification Standardizes AI Agent Tool Discovery
Google, Microsoft, and an industry coalition launched the Apache 2.0-licensed Agentic Resource Discovery specification to decentralize runtime tool routing.
Agentic Resource Discovery · Interoperability · Open Source
Ai Agents
Two-Agent AMIE Architecture Matches Physicians on 3-Visit Plans
Google Research demonstrated that its Gemini-based AMIE system matches primary care physicians in managing longitudinal patient care across multiple visits.
Google Research · Medical Ai · Longitudinal Care
Ai Engineering
Ai2's 4B MolmoMotion Maps Text Instructions to 3D Trajectories
Ai2 released MolmoMotion, an open-source 4B parameter model that predicts precise 3D physical trajectories from RGB video and natural language.
Open Source Models · 3d Motion Forecasting · Computer Vision
Ai Agents
MIT-Licensed GLM-5.2 MoE Reaches 74.4% on FrontierSWE
Zhipu AI has released GLM-5.2, a 744-billion parameter MoE model engineered for long-horizon agentic tasks with a stable one-million-token context window.
Mixture Of Experts · Open Source Ai · Long Context Window
Ai Agents
$3.6B Fin Acquisition Brings Verification-First AI to Agentforce
Salesforce has acquired autonomous customer service platform Fin in a $3.6 billion all-cash deal to integrate its reasoning engine into Agentforce.
Autonomous Agents · Customer Service Ai · Enterprise Ai
Ai Coding
15 JetBrains Plugins Caught Exfiltrating AI API Keys
Cybersecurity researchers at Aikido Security discovered a coordinated malware campaign using 15 JetBrains plugins to steal AI API keys from 70,000 developers.
Cybersecurity · Jetbrains Plugins · Api Security
Ai Agents
Alibaba's Qwen-Robot Suite Hits 45% Success on RoboChallenge
Alibaba's Qwen team has launched the Qwen-Robot Suite, a trio of foundation models targeting navigation, manipulation, and physical world simulation.
Alibaba Qwen · Robotics Foundation Models · Embodied Ai
Ai Engineering
$9M Seed Backs Probably's Deterministic AI Validation Layer
San Francisco startup Probably has raised $9 million from a16z and Accel to build a local validation layer that forces weaker LLMs to achieve 99.99% accuracy.
Deterministic Ai · Validation Layer · Venture Capital
Ai Agents
Domain Experts Sweep Claude Opus 4.7 Hackathon Results
Anthropic's latest hackathon highlights a shift in AI development, with doctors and teachers using Opus 4.7 to build complex agentic applications.
Anthropic · Claude Opus · Agentic Ai