AI News
E3 benchmark says agents waste tokens on simple tasks
A July 14 paper says estimate-first agent execution can match success while cutting cost, tokens, and inspected files on simple coding tasks.
By The AI Feed Desk
NVIDIA uses Japan to package physical AI as a full-stack ecosystem
NVIDIA's July 15 Japan ecosystem update ties RTX Spark, robotics, manufacturing, and local partners into a physical AI stack.
By The AI Feed Desk
Future of Life Institute grades nine AI labs on safety
The Summer 2026 AI Safety Index gives Anthropic a C+, OpenAI and Google DeepMind Cs, and failing grades to xAI, DeepSeek, and Mistral.
By The AI Feed Desk
GitHub moves Copilot security review and IDE trust controls into preview
GitHub's July 14 Copilot updates add app security review, Visual Studio MCP trust checks, C++ modernization GA, and JetBrains BYOK controls.
By The AI Feed Desk
Google brings Nano Banana image generation into Search AI Overviews
Google's 25th-anniversary Images update adds a personalized image home and brings Nano Banana image generation into AI Overviews.
By The AI Feed Desk
HCRMap targets hot-expert bottlenecks in MoE inference
A July 13 paper proposes HCRMap, a pressure-aware residency system for hot MoE experts across 3.5D chiplet memory tiers.
By The AI Feed Desk
AgentCheck tests whether MCP agents fail safely
A July 13 paper introduces AgentCheck, an MCP workbench for replaying tool faults and confirming whether agent mitigations actually fix them.
By The AI Feed Desk
Demis Hassabis calls for a US-led frontier AI watchdog
Google DeepMind CEO Demis Hassabis proposed a US-led standards body to test frontier AI models before release, including open and closed systems.
By The AI Feed Desk
Apple opens the public beta path for Siri AI
Apple's iOS 27 public beta path puts Siri AI in front of testers, shifting the story from WWDC promise to device, region, and workflow reality.
By The AI Feed Desk
FTC proposes AI accuracy-suppression policy statement
The FTC opened comments on a proposed policy statement that treats undisclosed AI output steering away from expected accuracy as potential deception under Section 5.
By The AI Feed Desk
NVIDIA and Hugging Face publish open data for agents
NVIDIA and Hugging Face published a Nemotron data package for agents, including open pretraining data, post-training samples, an interactive Prompt Atlas, and synthetic persona datasets.
By The AI Feed Desk
GitHub adds Codex as a JetBrains Copilot agent provider
GitHub added Codex as a public-preview agent provider in JetBrains Copilot, alongside hooks, MCP server management, approval modes, Claude permissions, debug logs, and custom models.
By The AI Feed Desk
OpenAI doubles bio-jailbreak bounty rewards for GPT-5.6
OpenAI turned its GPT-5.5 Bio Bug Bounty into an ongoing private Bio Bounty Program and raised the universal jailbreak reward to $50,000 for GPT-5.6 and GPT-5.5.
By The AI Feed Desk
ChatGPT Work turns ChatGPT into a desktop and app agent
OpenAI launched ChatGPT Work as a GPT-5.6-powered agent that can work across apps, files, browser tasks, scheduled tasks, Codex, documents, sheets, slides, and Sites.
By The AI Feed Desk
OpenAI publishes national-security AI partnership principles
OpenAI published principles for government and national-security partnerships as it expands defensive cyber and biosecurity work with allied governments.
By The AI Feed Desk
GitHub adds Copilot controls for settings and telemetry
GitHub added device-level Copilot managed settings and enterprise-managed OpenTelemetry export for VS Code and Copilot CLI.
By The AI Feed Desk
Anthropic and UST bring Claude into physical AI validation
Anthropic says UST is integrating Claude Code into chip validation and training 20,000 engineers, architects, and consultants worldwide.
By The AI Feed Desk