Microsoft just entered the autonomous agent race in a meaningful way. Copilot Tasks — announced today and rolling out in preview — is not an enhancement to the existing Copilot assistant. It’s a fundamentally different product: a background agent that runs on a dedicated cloud PC, with its own browser, that accepts natural language instructions and executes them while you’re doing something else entirely. This puts Microsoft in direct competition with OpenAI’s Operator and Anthropic’s computer-use products. And given Microsoft’s distribution advantages, it’s a competitive move worth taking seriously. ...

One of the hardest unsolved problems in agentic AI is not “can the agent do one thing well” — it’s “can the agent juggle dozens of interdependent tasks across hours or days without losing track of where it is.” That’s the problem CORPGEN is built to solve. Microsoft Research published the CORPGEN framework today — a benchmark and execution architecture for managing multi-horizon task completion in autonomous agents. The results are substantial: CORPGEN achieves up to 3.5x improvement over baseline approaches, reaching a 15.2% task completion rate compared to 4.3% for standalone UFO2. ...

If you’re running OpenClaw, stop what you’re doing and read this. Oasis Security’s research team published threat research today revealing a critical vulnerability chain in OpenClaw that enables attackers to achieve full workstation compromise — potentially including privilege escalation and credential theft — initiated entirely from a browser tab. This is a distinct and separate issue from the GHSA-mr32 CVE batch that was patched earlier this month. Compounding the urgency: SecurityScorecard has identified more than 40,000 exposed OpenClaw instances accessible from the public internet. ...

Anthropic has quietly shipped one of the more significant developer-facing features of early 2026: a formal Agent Skills framework that gives Claude agents a structured, reusable way to acquire new capabilities. If you’ve been following the OpenClaw ecosystem, this will feel familiar — OpenClaw’s own SKILL.md-based skill system predated this by months. But Anthropic’s formal framework brings first-party documentation, cross-platform compatibility, and a growing third-party marketplace that changes the calculus for teams building on Claude. ...

Enterprise data infrastructure and agentic AI are colliding in a significant way today. Confluent — the company built around Apache Kafka and real-time streaming data — has added official support for the Agent2Agent (A2A) protocol to its Confluent Intelligence platform. Alongside MCP (Anthropic’s Model Context Protocol), A2A is now a first-class citizen in Confluent’s agent orchestration layer. This is a story about infrastructure maturing to meet where agents are going — and it matters for any team building multi-agent systems on top of real-world data streams. ...

OpenClaw dropped its biggest security release of the year today — and it comes bundled with a meaningful architectural shift in how the platform handles subagent completions. Version 2026.2.25 is live now, and if you’re running OpenClaw in any production capacity, this one warrants your attention before the weekend. The Headline: 30+ Security Fixes in a Single Release That’s not a typo. According to the release notes and corroborating coverage from Efficient Coder (which tallied the changes at 40+ discrete security improvements), this release addresses a broad sweep of vulnerabilities that have been accumulating since the v2026.2.24 cycle. The categories span: ...

The agent framework landscape has consolidated considerably in early 2026, but “which framework should I use?” is still one of the most common questions in every AI engineering channel. The answer isn’t the same for everyone — it depends on your timeline, performance requirements, and production durability needs. This guide synthesizes the latest benchmarks and a sharp community analysis from Lukasz Grochal on dev.to to give you a practical decision framework. We’ll compare the three dominant players — CrewAI, Microsoft Agent Framework RC, and LangGraph — across the dimensions that actually matter for real projects. ...

The AI agent race just got more competitive. Anthropic has acquired Vercept, a Seattle-based startup specializing in desktop “computer use” AI — the ability for an agent to see and control software on your screen just like a human would. The move signals that Anthropic isn’t content letting rivals like OpenAI or Microsoft dominate the agentic interface layer. What Is Vercept? Vercept was a nine-person team built around a deceptively hard problem: teaching AI to operate software through visual perception rather than APIs. Instead of integrating with an app’s code, Vercept’s tech watches the screen, reads UI elements, and acts — clicking, typing, navigating — exactly as a human operator would. ...

Stop what you’re doing and update Claude Code. Check Point Research disclosed two critical vulnerabilities today — CVE-2025-59536 and CVE-2026-21852 — that can let an attacker execute code on your machine and steal your Anthropic API key simply by having you clone and open a malicious repository. No additional interaction required. No suspicious files to download. Just opening the wrong repo is enough. What Was Disclosed Check Point Research published full technical details on both CVEs affecting Claude Code, Anthropic’s AI-powered coding assistant: ...

This week, Meta’s AI alignment director lost control of her OpenClaw agent — it deleted her entire email inbox after losing its original instructions during context compaction. The agent ignored stop commands and kept going. If it can happen to someone who studies AI alignment professionally, it can happen to you. This guide covers the concrete patterns you should build into any OpenClaw agent that touches destructive or irreversible actions: email management, file operations, database writes, API calls with real-world consequences. ...