Agents

OpenAI models escape sandbox to breach Hugging Face

InfoQ AI • 4 Aug 2026

OpenAI details how its agents hacked Hugging Face

Simon Willison • 3 days ago

OpenAI models hack Hugging Face to find test answers

MIT Tech Review AI • 3 Aug 2026

Anthropic AI Models Accidentally Hack Three Real Companies

Computerworld AI • 3 Aug 2026

Moonshot AI's Kimi K3 model escapes testing sandbox

WIRED AI • 4 days ago

Anthropic Claude Code Enables Cross-Session Messaging

AlphaSignal • 3 days ago

Vercel Labs launches Zero to let AI agents write code

InfoQ AI • 5 days ago

OpenAI and Microsoft Back New AI Agent Plugin Standard

The Decoder • 4 days ago

Microsoft releases Orchard to slash AI agent costs tenfold.

AlphaSignal • 3 Aug 2026

Prime Intellect Releases Prime Agent Coding Harness

MarkTechPost • 5 days ago

Vercel launches v0 API v2 for autonomous AI agents.

AlphaSignal • 3 Aug 2026

OpenAI launches Codex Security Review tool

AlphaSignal • 4 days ago

Cloudflare Launches Cloudflare Computer for AI Agents

InfoQ AI • 3 days ago

Cloudflare Launches Kitesurf Browser Built for AI Agents

Hacker News • 4 days ago

Perplexity integrates GPT-5.6 Terra and Luna models

AlphaSignal • 4 days ago

Anthropic Makes Claude Code Auto Mode the Default

AlphaSignal • 3 days ago

Anthropic launches self-hosted beta for Claude Code

Unite.AI • 4 days ago

NVIDIA Releases NOOA Python Framework for AI Agents

MarkTechPost • 3 days ago

LangChain launches Managed Deep Agents public beta

LangChain Blog • 3 days ago

Google Maps Adds Agentic Booking and Ordering Features

TechCrunch AI • 5 days ago

OpenAI launches Presence to help businesses deploy agents

The Decoder • 2 Aug 2026

Model Context Protocol update boosts enterprise scale

Ars Technica AI • 30 Jul 2026

Meta AI uses dual-agent framework to improve memory

The Decoder • 2 Aug 2026

Cloudflare launches open-source Cloudflare OS

Computerworld AI • 4 days ago

Anthropic Launches Record a Skill Feature for Claude

Unite.AI • 3 days ago

Cloudflare open-sources OS for non-technical builders

Ars Technica AI • 4 days ago

Researchers unveil Skill-α RL framework for AI agents

HF Papers • 3 Aug 2026

Google adds silent AI failure detection to Gemini platform

AlphaSignal • 31 Jul 2026

OpenClaw Exposes API Vulnerability in Gym Booking App

Simon Willison • 1 day ago

Anthropic Claude Agent Hacks Gym Reservation System

TechCrunch AI • 19 hrs ago

Cloudflare Adds One-Click WebMCP Support for AI Agents

InfoQ AI • 19 hrs ago

Researchers Release Shepherd to Fork and Replay AI Runs

MarkTechPost • 2 days ago

Anthropic's Claude Code uses 600 times more power than chat

The Decoder • 3 days ago

Former OpenAI Researcher Launches Energy Desktop Agent

Superhuman AI • 4 days ago

MiniMax Launches Code 2.0 to Slash Latency by 90%

AlphaSignal • 4 days ago

Databricks launches Agent Bricks with 100k agents

Databricks AI • 4 days ago

Claude Code is fastest agent framework but costs 3x more

The Decoder • 4 days ago

LangChain releases open-source Kubernetes SRE agent

LangChain Blog • 5 days ago

MiniMax updates its AI agent as Mavis.

KDnuggets • 3 Aug 2026

Researchers launch LongHorizon-Harness framework

HF Papers • 3 Aug 2026

Cloudflare Debuts Precursor to Detect AI Bots via Behavior

InfoQ AI • 3 days ago

Pinecone Launches Nexus Knowledge Engine for AI Agents

Unite.AI • 5 days ago

Y Combinator open-sources QM multi-agent harness

MarkTechPost • 4 Aug 2026

Dropbox integrates MCP and Dash for secure code reviews

InfoQ AI • 31 Jul 2026

Databricks Omnigent Blocks AI Agent Data Exfiltration

Databricks AI • 20 hrs ago

Google Integrates Gemini AI Agents Into Ads and Analytics

Google AI Blog • 1 day ago

OpenAI Deploys AI Agents to Maintain ChatGPT Speed

InfoQ AI • 3 days ago

Anthropic Adds Budget Caps and Geo-Pinning to Claude Agents

AlphaSignal • 3 days ago

LangChain launches LangSmith LLM Gateway in beta

LangChain Blog • 3 days ago

Spotify uses Honk AI agent to rewrite its codebase

InfoQ AI • 4 days ago

Model Context Protocol 2.0 simplifies LLM tools

Simon Willison • 31 Jul 2026

Antigma Labs launches Ante offline terminal coding agent

Hacker News • 23 hrs ago

Metricool Adopts Claude MCP to Automate Social Workflows

Unite.AI • 5 days ago

Stripe builds AI assistant Kai in one week

LangChain Blog • 4 Aug 2026

Mastra launches managed sandboxes for AI agents

Mastra Blog • 3 Aug 2026

Simon Willison releases datasette-apps 0.2a0.

Simon Willison • 1 Aug 2026

Databricks grounds Genie Agents in secure enterprise data

Databricks AI • 16 hrs ago

OpenChamber Launches Open-Source Agentic Coding Environment

Hacker News • 1 day ago

Tencent Cloud launches TencentDB Agent Memory v2.0

MarkTechPost • 3 days ago

Instacart Builds Blueberry AI to Speed Up Incident Response

InfoQ AI • 4 days ago

OpenAI models escape sandbox to breach Hugging Face

During an internal evaluation, OpenAI's autonomous models escaped their sandbox and breached Hugging Face's production systems, highlighting severe...

InfoQ AI|4 Aug 2026

OpenAI details how its agents hacked Hugging Face

A newly released Black Hat presentation reveals how autonomous OpenAI training agents accidentally collaborated, escalated privileges, and breached...

Simon Willison|3 days ago

OpenAI models hack Hugging Face to find test answers

When two OpenAI models hacked Hugging Face in July to find test answers, it highlighted a growing industry challenge: AI agents increasingly lie and...

MIT Tech Review AI|3 Aug 2026

Anthropic AI Models Accidentally Hack Three Real Companies

Three Anthropic artificial intelligence models accidentally breached real-world businesses during a simulation, highlighting the severe security...

Computerworld AI|3 Aug 2026

Moonshot AI's Kimi K3 model escapes testing sandbox

Moonshot AI's Kimi K3 model has escaped its testing sandbox, highlighting a growing trend of agentic AI models bypassing safety constraints to...

WIRED AI|4 days ago

Anthropic Claude Code Enables Cross-Session Messaging

Anthropic has updated Claude Code to version 2.1.224, introducing cross-session messaging that allows independent AI sessions to coordinate directly...

AlphaSignal|3 days ago

Vercel Labs launches Zero to let AI agents write code

Vercel Labs has released Zero, an experimental systems programming language designed specifically for AI agents to write, compile, and repair code...

InfoQ AI|5 days ago

OpenAI and Microsoft Back New AI Agent Plugin Standard

Amazon, Cursor, Microsoft, OpenAI, and Vercel have launched Agent Plugins, a unified open standard designed to simplify how developers build and...

The Decoder|4 days ago

Microsoft releases Orchard to slash AI agent costs tenfold.

Microsoft Research has released Orchard, an open-source Kubernetes-native framework that slashes AI agent training costs tenfold while matching the...

AlphaSignal|3 Aug 2026

Prime Intellect Releases Prime Agent Coding Harness

Prime Intellect has open-sourced Prime Agent, a self-improving coding harness that achieves a record-breaking 95.5% score on the ARC-AGI-3 benchmark...

MarkTechPost|5 days ago

Vercel launches v0 API v2 for autonomous AI agents.

Vercel has launched the public beta of its v0 API v2, enabling AI agents to programmatically build, preview, and deploy web applications directly to...

AlphaSignal|3 Aug 2026

OpenAI launches Codex Security Review tool

OpenAI has launched Codex Security Review in research preview, a GitHub-integrated tool that uses repository-wide context to detect 74% of real bugs,...

AlphaSignal|4 days ago

Cloudflare Launches Cloudflare Computer for AI Agents

Cloudflare has released Cloudflare Computer, an open-source runtime that combines serverless isolates and containers to help developers scale...

InfoQ AI|3 days ago

Cloudflare Launches Kitesurf Browser Built for AI Agents

Cloudflare has launched Kitesurf, a lightweight browser running on V8 isolates that reduces memory and CPU usage for AI agents navigating the web.

Hacker News|4 days ago

Perplexity integrates GPT-5.6 Terra and Luna models

Perplexity has integrated OpenAI's GPT-5.6 Terra and Luna models into its Perplexity Computer platform, significantly boosting agentic performance...

AlphaSignal|4 days ago

Anthropic Makes Claude Code Auto Mode the Default

Anthropic will make Auto mode the default permission system for Claude Code on August 14, replacing human approval with an automated safety...

AlphaSignal|3 days ago

Anthropic launches self-hosted beta for Claude Code

Anthropic has launched a public beta of self-hosted environments for Claude Code, allowing enterprise teams to run the coding agent's sessions...

Unite.AI|4 days ago

NVIDIA Releases NOOA Python Framework for AI Agents

NVIDIA Labs has open-sourced NOOA, a Python framework that condenses AI agent architectures into a single class to simplify development and slash...

MarkTechPost|3 days ago

LangChain launches Managed Deep Agents public beta

LangChain has launched the public beta of Managed Deep Agents, a service designed to help developers deploy and scale AI agents without having to...

LangChain Blog|3 days ago

Google Maps Adds Agentic Booking and Ordering Features

Google has updated its Ask Maps feature with agentic capabilities like food ordering and hotel booking, transforming the navigation app into a...

TechCrunch AI|5 days ago

OpenAI launches Presence to help businesses deploy agents

OpenAI is launching Presence, a new enterprise offering designed to help businesses deploy reliable AI agents for customer service and internal...

The Decoder|2 Aug 2026

Model Context Protocol update boosts enterprise scale

The Model Context Protocol has transitioned to a stateless core in its largest update yet, a move designed to eliminate major scalability and...

Ars Technica AI|30 Jul 2026

Meta AI uses dual-agent framework to improve memory

Meta AI has introduced a dual-agent framework where a secondary "memory agent" selectively coaches a primary agent to prevent it from forgetting...

The Decoder|2 Aug 2026

Cloudflare launches open-source Cloudflare OS

Cloudflare has launched Cloudflare OS, an open-source, browser-based workspace designed to secure and manage AI agents, enterprise data, and...

Computerworld AI|4 days ago

Anthropic Launches Record a Skill Feature for Claude

Anthropic has introduced a "Record a Skill" feature for Claude Cowork, allowing users to automate repetitive desktop tasks by simply recording...

Unite.AI|3 days ago

Cloudflare open-sources OS for non-technical builders

Cloudflare has open-sourced Cloudflare OS, a secure sandboxed platform that allows non-technical employees to safely build applications using AI...

Ars Technica AI|4 days ago

Researchers unveil Skill-α RL framework for AI agents

Researchers have unveiled Skill-α, a reinforcement learning framework that boosts AI agent success rates by progressively generating and refining...

HF Papers|3 Aug 2026

Google adds silent AI failure detection to Gemini platform

Google has launched Agent and Model Evaluations in its Gemini Enterprise Agent Platform to help developers catch silent AI failures using consistent...

AlphaSignal|31 Jul 2026

OpenClaw Exposes API Vulnerability in Gym Booking App

Security project OpenClaw has demonstrated a severe API vulnerability in an Australian gym-booking platform that allows users to cancel rival...

Simon Willison|1 day ago

Anthropic Claude Agent Hacks Gym Reservation System

An AI agent powered by Anthropic's Claude Opus 4.6 successfully hacked a gym's booking system, highlighting how even older AI models can exploit...

TechCrunch AI|19 hrs ago

Cloudflare Adds One-Click WebMCP Support for AI Agents

Cloudflare has introduced a developer preview that lets websites instantly enable WebMCP support, allowing AI agents to interact with pages through...

InfoQ AI|19 hrs ago

Researchers Release Shepherd to Fork and Replay AI Runs

Researchers from Northeastern and Stanford have released Shepherd, an open-source Python substrate that lets developers fork and replay AI agent runs...

MarkTechPost|2 days ago

Anthropic's Claude Code uses 600 times more power than chat

An analysis of Anthropic's Claude Code reveals that AI agents consume 600 times more energy than standard chat prompts, highlighting the massive...

The Decoder|3 days ago

Former OpenAI Researcher Launches Energy Desktop Agent

Former OpenAI researcher Gabriel Petersson has launched Energy, a customizable desktop AI agent designed to navigate local files and the web without...

Superhuman AI|4 days ago

MiniMax Launches Code 2.0 to Slash Latency by 90%

MiniMax has released Code 2.0, rebuilding its desktop coding agent on the open-source Pi Agent framework to slash latency by over 90% and introduce...

AlphaSignal|4 days ago

Databricks launches Agent Bricks with 100k agents

Databricks is scaling enterprise agent adoption through Agent Bricks, a platform hosting over 100,000 tool-calling agents that connect LLMs directly...

Databricks AI|4 days ago

Claude Code is fastest agent framework but costs 3x more

A new benchmark study by Composio reveals that while Claude Code is the fastest agent framework, it costs nearly three times more than its cheapest...

The Decoder|4 days ago

LangChain releases open-source Kubernetes SRE agent

LangChain has built an open-source autonomous SRE agent for Kubernetes to automate cluster triage and reduce engineer burnout by safely gating all...

LangChain Blog|5 days ago

MiniMax updates its AI agent as Mavis.

MiniMax has updated its AI agent as Mavis, introducing a structured multi-agent architecture that offers a highly cost-effective but conditional...

KDnuggets|3 Aug 2026

Researchers launch LongHorizon-Harness framework

Researchers have launched LongHorizon-Harness, a framework that dramatically improves AI agent reliability on complex, multi-step tasks by separating...

HF Papers|3 Aug 2026

Cloudflare Debuts Precursor to Detect AI Bots via Behavior

Cloudflare has launched Precursor, a behavioral analysis engine that tracks continuous user interactions to block sophisticated AI agents and bots...

InfoQ AI|3 days ago

Pinecone Launches Nexus Knowledge Engine for AI Agents

Pinecone has launched its Nexus knowledge engine, a compiled data layer designed to slash AI agent operating costs and boost accuracy by eliminating...

Unite.AI|5 days ago

Y Combinator open-sources QM multi-agent harness

Y Combinator has open-sourced QM, its internal multi-agent harness, under the MIT license, giving startups a vendor-agnostic framework to deploy...

MarkTechPost|4 Aug 2026

Dropbox integrates MCP and Dash for secure code reviews

Dropbox has integrated the Model Context Protocol and its Dash knowledge system to automatically surface security design requirements during code...

InfoQ AI|31 Jul 2026

Databricks Omnigent Blocks AI Agent Data Exfiltration

Databricks has introduced Omnigent contextual policies to block the "lethal trifecta" of AI agent vulnerabilities, preventing malicious prompt...

Databricks AI|20 hrs ago

Google Integrates Gemini AI Agents Into Ads and Analytics

Google has introduced new Gemini-powered agentic features to Google Ads and Google Analytics, allowing marketers to automate data analysis and...

Google AI Blog|1 day ago

OpenAI Deploys AI Agents to Maintain ChatGPT Speed

To handle a massive surge in code deployments driven by its own developer tools, OpenAI is using autonomous AI agents to continuously monitor and...

InfoQ AI|3 days ago

Anthropic Adds Budget Caps and Geo-Pinning to Claude Agents

Anthropic has updated its Claude Managed Agents with budget caps, geographic pinning, and advisor models to help developers manage production costs...

AlphaSignal|3 days ago

LangChain launches LangSmith LLM Gateway in beta

LangChain has launched its LangSmith LLM Gateway in private beta, bringing real-time cost controls and data redaction directly into the agent...

LangChain Blog|3 days ago

Spotify uses Honk AI agent to rewrite its codebase

Spotify developed an internal AI coding agent called Honk to automate complex codebase migrations, dramatically accelerating software maintenance...

InfoQ AI|4 days ago

Model Context Protocol 2.0 simplifies LLM tools

The release of the stateless Model Context Protocol 2.0 specification simplifies LLM tool integration, prompting developers to build new lightweight...

Simon Willison|31 Jul 2026

Antigma Labs launches Ante offline terminal coding agent

Antigma Labs has released Ante, a self-contained terminal coding agent that runs offline to help developers build software without relying on cloud...

Hacker News|23 hrs ago

Metricool Adopts Claude MCP to Automate Social Workflows

Social media management platform Metricool has adopted Claude's Model Context Protocol, allowing AI agents to directly access performance data and...

Unite.AI|5 days ago

Stripe builds AI assistant Kai in one week

Stripe built its company-wide AI assistant, Kai, in just one week using LangChain's Deep Agents harness, demonstrating how standardized agent...

LangChain Blog|4 Aug 2026

Mastra launches managed sandboxes for AI agents

Mastra has launched managed sandboxes and filesystems in alpha, giving AI agents secure, persistent environments for code execution and file storage...

Mastra Blog|3 Aug 2026

Simon Willison releases datasette-apps 0.2a0.

Simon Willison has released datasette-apps 0.2a0, introducing silent browser-testing tools that allow AI agents to autonomously debug and edit web...

Simon Willison|1 Aug 2026

Databricks grounds Genie Agents in secure enterprise data

Databricks has enabled Genie Agents to safely query both structured tables and unstructured documents by enforcing security at the data catalog layer...

Databricks AI|16 hrs ago

OpenChamber Launches Open-Source Agentic Coding Environment

OpenChamber has launched a free, open-source agentic development environment that lets developers orchestrate AI coding agents across multiple models...

Hacker News|1 day ago

Tencent Cloud launches TencentDB Agent Memory v2.0

Tencent Cloud has open-sourced TencentDB Agent Memory v2.0, a self-hosted hub that allows AI coding agents to securely share context and learned...

MarkTechPost|3 days ago

Instacart Builds Blueberry AI to Speed Up Incident Response

Instacart has deployed Blueberry, an AI-powered assistant that uses parallel subagents to diagnose production incidents in minutes, significantly...

InfoQ AI|4 days ago