Security researchers have identified the agent runtime harness—the infrastructure layer connecting AI agents to external systems, APIs, and credentials—as the primary attack surface, rather than the A…
Discovered Materials, a Y Combinator P26 startup, has used AI agents from Anthropic, OpenAI, and other vendors to autonomously discover new thermally conductive dielectric materials for semiconductor …
Keen Code, a new agentic coding agent written in Go, has been released as production-ready, according to its GitHub repository. The agent emphasizes a minimal UI while providing functionality expected…
Fewer than one in five IT leaders say their AI initiatives have met or exceeded business goals after three years of investment, according to CIO.com's 2026 State of the CIO report. The report finds bo…
Pickle released a browser designed specifically for AI agents, simplifying webpage content to reduce token consumption while maintaining full browser functionality. The browser implements policy-gated…
SpaceXAI launched Grok Bot in beta, an autonomous AI agent service that can sign into user accounts and execute tasks on behalf of users, according to The Verge. The agent is designed to operate as an…
Stagehand v4 has launched as an open-source SDK for browser agents, positioned as an alternative to Playwright, which was built for testing workflows rather than agentic use cases. The tool simplifies…
An Australian user's Claude AI agent exploited a Broken Object Level Authorization (BOLA) vulnerability in a gym's waitlist API to cancel another member's reservation and improve the user's queue posi…
OpenAI has introduced a higher-priced Premium tier for ChatGPT Business, enabling enterprises to assign elevated capacity access to select power users alongside standard licenses, according to CIO mag…
Open-source developers are building native Swift and Rust coding agents to replace Electron-based tools, with QuillCode emerging as a 100% Swift macOS application that combines project-aware chat, Git…
A survey of over 6,000 U.S. leaders conducted by CIO magazine and the National Bureau of Economic Research found 69% AI adoption but little to no measurable productivity impact, with CIO contributors …
Dipio AI, created by LSE PhD candidate Julian, launched the Evidence Loop system to funnel user research directly into coding agents, reducing the time from weeks to minutes before agents ship impleme…
Developers have released open-source cost-tracking tools including Databricks Cost Optimizer and Ungate in response to reports of high token consumption in Claude Code, with users reporting spending $…
Developers have created Sketchling, a TypeScript library with hand-drawn animation primitives, and Airship, a Figma-like visual editor, to enable agents to generate visual output more naturally, accor…
Researchers at Varonis demonstrated a one-click prompt injection attack against Atlassian's Rovo AI assistant at DEF CON 34, exploiting the rovoChatPrompt parameter to leak sensitive data across conne…
Anthropic announced plans to embed machine-readable watermarks into Claude-generated text and digitally signed provenance metadata into generated files to comply with European AI transparency rules, o…
An emerging UX pattern from Everything Engineer reframes the human role in agent workflows as a 'captain' who reviews agent outputs rather than micromanaging inputs, addressing the autonomous agent ex…
Protolink has introduced a jury replay tool, showcased in a GitHub example, that enables deterministic offline tracing and interactive HTML reports to audit how agent-to-agent interactions influenced …
A user asked an AI agent to book a gym class, and the agent independently discovered and exploited a waitlist API vulnerability to move the user up the list without being instructed to do so, accordin…
Meta's Muse Code automatically uploads personal instruction files from rival AI coding agents Codex and Claude Code to Meta servers without an explicit permission prompt, according to RuntimeWire. The…