Autonomous AI agents, agentic workflows, tool use, computer use, and multi-agent systems.
1,407 articles
The second day of Berkeley RDI's Agentic AI Summit 2026 covered topics ranging from governance and evaluations to research, platforms, finance and startups. The recap highlights the shift from oversight discussions toward practical discovery and deployment. Source: medium.com
In early August 2026, Varonis Systems launched Agent Intent-Based Access Control within its Atlas platform, adding new safeguards to monitor what AI agents are trying to do. The announcement, paired with an improved 2026 outlook, sent shares up nearly 9%. Source: simplywall.st
Climate scientist Zeke Hausfather tracked his own Claude Code usage over eight weeks, logging 3.2 billion tokens and roughly 170 kWh of data center electricity. The findings underscore how much more energy-intensive agentic AI tasks are compared to basic chatbot prompts. Source: the-decoder.com
Kiteworks, which helps organizations manage risk around sharing and handling private data, has announced a strategic partnership with Reco. The collaboration aims to improve monitoring and data governance around AI agent activity. Source: cybersecurity-insiders.com
Following up on a previous project, the author describes building a Streamlit-based UI for an AI agent created with LangGraph. The agent was originally designed to automate a 15-minute customer service booking workflow. Source: towardsdatascience.com
Salesforce reports that companies are deploying AI agents into production faster than before, based on data from its customer base. The growth in agentic AI usage has more than doubled compared to the previous year. Source: cio.com
On August 4, 2026, the US Ninth Circuit Court of Appeals struck down a preliminary injunction that had prevented Perplexity's AI agent from accessing certain third-party websites. The ruling addresses how the Computer Fraud and Abuse Act applies to autonomous AI agents browsing the web. Source: jdsupra.com
The author notes that most AI agent interfaces still look the same - either a chat window awaiting a reply or a CLI streaming raw structured output. The Agent Terrarium project aims to explore a fresh way to visualize and interact with agents. Source: commandline.microsoft.com
Attackers used typosquatted names mimicking popular AI services Paperclip and Browser Use to push malicious skills onto skills.sh's trending list. These fake skills instructed AI agents to install additional harmful components. Source: csoonline.com
Agentic AI security involves protecting autonomous systems as they plan, reason, access data, and take actions using various tools across enterprise environments. The article outlines key practices for reducing risks tied to human oversight of these agents. Source: proofpoint.com
nCino, a platform for agentic AI in banking, has launched new functionality letting lenders directly connect MCP-compatible AI agents to its system. The move aims to streamline mortgage workflows through agent integration. Source: fintech.global
As AI agents become more capable of completing online tasks, developers face the question of how sites can support them without relying on clicking buttons or filling forms manually. The article explains WebMCP as an emerging standard designed for this purpose. Source: sitepoint.com
Jim Liddle, previously Chief Innovation Officer at Nasuni, has formed strong views on how companies should implement agentic AI and deploy so-called digital employees. His perspective draws on his experience overseeing AI strategy at the company. Source: blocksandfiles.com
The VA, Air Force, and Department of Labor are rolling out AI agents to automate internal workflows and improve citizen-facing services. The report outlines how agencies can start building similar agentic systems. Source: fedtechmagazine.com
Microsoft explains how it uses its own Agent 365 platform internally to oversee and manage AI agents across the company. The company frames this as part of its shift toward becoming a so-called Frontier Firm. Source: microsoft.com
Liquid AI has released LFM2.5-2.6B, a 2.69-billion-parameter model built to run directly on devices. It supports a 128K context window, tool calling, and comes with open weights. Source: marktechpost.com
Startup Liquid AI, founded in 2023 by former MIT researchers, has released LFM2.5-2.6B, an open-weight language model designed to run without cloud servers or GPUs. It's small enough to power AI agents on hardware as limited as a Raspberry Pi. Source: venturebeat.com
A new system called Dogwood aims to verify AI agents' actions while they run, focusing on tool calls that let agents interact with the outside world. These tool calls are also flagged as a key source of risk in agentic systems. Source: aws.amazon.com
A developer describes building a custom AI data agent capable of querying enterprise data and answering business questions directly. The project follows up on an earlier piece about the challenges of creating AI-native enterprise data platforms. Source: towardsdatascience.com
A guide argues that agentic AI will be the dominant AI theme of 2026, but notes most existing coverage focuses on business productivity rather than technical implementation. It aims to fill that gap with developer-oriented explanations of autonomous agents. Source: sitepoint.com