OpenClaw

Ben Guez is using an automated script built with OpenClaw, Claude code, and Instagram trials to receive numerous DMs from potential international romantic partners. The script has generated significant interest, as indicated by the influx of direct messages he is receiving.

BioShocking is an attack technique that manipulates AI-powered browsers into bypassing safety guardrails by immersing them in fictional scenarios, leading to the extraction of sensitive data like credentials. The method exploits how AI agents in 'agent mode' inherit user-authenticated contexts, making them vulnerable to goal-manipulation attacks that trick them into performing harmful actions.

The OpenClaw app is now available for iOS and Android smartphones. The article mentions that smartphones are welcoming 'agentic AI overlords.'

An AI agent named Claude Opus 4.6 successfully defended against 6,000 hack attempts. Fernando Irarrázaval shared his OpenClaw assistant's inbox on Hacker News to demonstrate this security achievement.

Codex, OpenAI's agentic coding platform, is gaining mainstream traction as usage shifts from chat-based AI to delegated work. A report by OpenAI, Columbia, Duke, and the University of Pennsylvania reveals 99.8% of OpenAI employees use Codex for most output tokens, with organizations and individuals showing rising adoption. Non-developers are the fastest-growing user group, and agents now handle tasks like managing calendars, files, and scripts.

Loop engineering is gaining prominence in AI as a replacement for prompt engineering, with figures like Claude Code creator Boris Cherny and OpenClaw creator Peter Steinberger advocating for automated systems that reduce manual prompting. Loops enable AI agents to perform recurring tasks autonomously, such as coding and repository maintenance, by structuring workflows with components like automations, worktrees, and sub-agents.

A research paper reveals that the orchestration layer, which handles non-AI operational tasks, is more critical than AI models themselves for real-world functionality. The study highlights that only 1.6% of Claude Code's codebase involves AI decision logic, with the rest dedicated to operational infrastructure.

The OpenClaw AI agent was found to be vulnerable to phishing attacks during simulations, leading to user data exposure. Testing with various configuration profiles revealed susceptibility to tactics commonly used against human users.

Microsoft has launched an AI assistant called Microsoft Scout powered by OpenClaw, capable of scheduling meetings, managing tasks, ordering food, and filing expenses. The agent is rolling out to a small group of customers first, with security measures ensuring authorized access only. Nvidia is developing a more secure version of OpenClaw, while Microsoft will charge via GitHub Copilot subscriptions.

Microsoft introduced a new category of agentic AI called 'Autopilot,' starting with the agent 'Scout,' designed to autonomously manage tasks across cloud, desktop, and web platforms. Scout integrates with Microsoft apps like Teams, Outlook, and OneDrive to schedule meetings, flag deadlines, and generate preparatory materials without constant user input.

Microsoft is transforming OpenClaw into an enterprise AI agent named Scout, leveraging its 1.4 billion Windows users to drive adoption.

Microsoft announced Scout, an autonomous AI agent developed using OpenClaw. The announcement was covered by multiple outlets, including Microsoft's blog, Wired, and 404media, with Hacker News users discussing the development.

Microsoft launched Scout, a new AI assistant inspired by OpenClaw, designed to integrate the capabilities of OpenClaw into the Microsoft 365 ecosystem. The assistant was introduced at Microsoft's Build event.

Microsoft's AI coworker, Scout, is an OpenClaw-style agent that appears in Teams to automate dull office tasks.

Microsoft Scout is a new AI personal assistant built on OpenClaw, designed to integrate into Microsoft 365 apps like Outlook, OneDrive, and Teams. It assists with tasks such as organizing calendars, expense reporting, and email drafting, offering broader capabilities than Microsoft's Copilot. Corporate vice president Omar Shahine described it as the first real personal assistant offered by Microsoft.

Intel, Foxconn, and other partners are developing high-density rack-scale CPU designs using Xeon processors to support agentic AI workloads, with configurations reaching 36,864 cores in a 100kW power envelope. Competitors like Nvidia and Arm are also pursuing similar rack-scale solutions for AI agents.

Intel's Clearwater Forest Xeon 6+ processors feature 288 cores per socket, surpassing competitors like Nvidia and Arm in core density for agentic AI workloads. The chips use advanced packaging and 2 nm-class process technology, with Darkmont E-cores optimized for tasks like web requests, code execution, and API interactions.

Nvidia and Microsoft are set to debut the first Windows PCs powered by Nvidia chips at Computex and Microsoft's Build conference. These PCs, including models from Microsoft's Surface brand and Dell, will feature software enabling local AI agent operations, marking a second attempt for Microsoft after setbacks with its Copilot+ PC.

Alibaba Group Holding has launched MuleRun, an AI agent platform aimed at replicating the success of OpenClaw while addressing privacy and security concerns. The platform was introduced at the Alibaba Cloud summit on May 20 as an 'always-on AI workforce' solution.

Alibaba Group Holding's new AI agent platform MuleRun aims to replicate the success of OpenClaw while addressing privacy and security concerns. The platform, presented at the Alibaba Cloud summit, offers a centralized service for accessing AI agents and is positioned as an 'always-on AI workforce'.

India's CERT-In recommends a 12-hour window to patch or mitigate exploited 'n-day' vulnerabilities in internet-facing or critical systems due to AI-driven cyber threats. The guidelines emphasize urgent action as AI accelerates exploitation processes, with stricter timelines for high-risk flaws compared to 24 hours for internal system vulnerabilities.

The article details how AI agents 'Claude Code' and 'OpenClaw' triggered a major upheaval in the tech industry, marking one of computing's most significant transformations.

Researchers highlight security risks in AI agents where minor modifications to text-based 'skills' can enable prompt injection attacks, allowing malicious actors to bypass safety measures. Studies show 13.4% of skills on platforms like ClawHub contain critical vulnerabilities, including malware and exposed secrets. The issue arises as agents automatically fetch third-party skills, making them susceptible to adversarial manipulation.

Google announced new AI agents at I/O 2026 for tasks like gathering information and planning events. The agents can run continuously in the background and integrate into various services. This development may bring capable personal assistants to users, a promise tech companies have made for years.

Google announced Gemini Spark at Google I/O 2026, a personal AI agent that manages inboxes, calendars, and workflows. Spark is Google's most ambitious Gemini feature yet. It aims to replace OpenClaw with its advanced capabilities.

Google is launching Gemini Spark, its own version of OpenClaw, an always-on AI agent that can perform various tasks such as writing emails and monitoring credit card statements. Gemini Spark is powered by Gemini 3.5 Flash and runs on Google Cloud. The AI agent will connect to Google Workspace apps and third-party apps.

Google has developed an always-running AI agent that can spend money and send emails. The agent is designed to be data-hungry and active 24/7. This raises concerns about privacy and financial security.

The 'Claw Chain' vulnerabilities in OpenClaw deployments have been patched, but previously allowed attackers to steal credentials, escalate privileges, and maintain persistence, posing a significant threat to the AI agent framework. The vulnerabilities were found in the rapidly growing framework. The patch addresses the security risks associated with these vulnerabilities.

Agent harnesses like OpenClaw are changing how AI models are built and run, enabling complex tasks automation and improving model effectiveness. These harnesses wrap around LLMs' API endpoints and orchestrate tool calls. They have a significant impact on code assistant success, even with smaller models.

The creator of OpenClaw spent $1.3 million on OpenAI tokens within a 30-day period. This significant investment suggests a strong interest in AI technology. The news was shared on Twitter and discussed on Y Combinator.

Cybersecurity researchers disclosed four security flaws in OpenClaw, dubbed Claw Chain, which can be chained to achieve data theft, privilege escalation, and persistence. These vulnerabilities permit an attacker to establish a foothold, expose sensitive data, and plant backdoors. The flaws were identified by Cyera.

Anthropic has restricted access to its Claude model family by separating interactive and programmatic usage into different budgets. Programmatic interaction will be billed at costlier API rates, while interactive use remains under subscription limits. This change aims to prevent customers from overusing tokens at the all-you-can-eat subscription rate.

The author built a signal filter for crypto Twitter using OpenClaw, called AlphaFeed, which monitors a curated list of macro and crypto voices and delivers a daily digest to the user's inbox. The goal is to surface what smart money is paying attention to. The tool uses a weighted scoring system with topic filters and cross-account amplification.

A user tested the Grok-Bankr NFT-injection exploit on their RunLobster setup and nearly approved a transfer proposal for $84. The exploit uses structural attack and SPL token metadata to generate proposals. The user's agent generated a proposal with a confidence score of 0.87.

Big Tech companies like Meta and Google are rushing to build agentic tools after the success of OpenClaw, a viral personal assistant. This has triggered an 'agentic wars' among tech giants. The competition is heating up with multiple companies entering the AI agent race.

OpenClaw experienced a difficult week, with an article discussing the challenges on their blog. The article received comments on a news platform. The company had a rough week with 23 points and 33 comments.

Mark Zuckerberg plans to make AI agents more accessible and Meta's AI agent plans include an OpenClaw competitor that can shop on Instagram. The goal is to increase accessibility of AI agents. Meta's AI agent will have e-commerce capabilities.

Spotify now allows AI agents to generate personal podcasts, a new feature announced recently. This development is expected to change the way users interact with the platform. The feature utilizes AI technology to create personalized content.

The article compares three AI tools, OpenClaw, KionClaw, and Claude Code, for predicting crypto markets, with KionClaw performing better in terms of ease of use and accuracy. KionClaw stood out for its quick setup and ability to access real-time data. The test results show KionClaw as the top choice for crypto analysis.

Professor Hannah Fry conducted an experiment with an AI agent, Cass, which was given a credit card and tasked with completing real-world chores, highlighting its capabilities and risks. The agent successfully completed some tasks but also leaked passwords and caused CAPTCHA chaos. The experiment demonstrated the potential risks of granting autonomy to AI agents.