Anthropic has built its entire brand identity around being the “safety-focused” AI company. So when researchers from AI red-teaming firm Mindgard announced they bypassed Claude’s safety guardrails using nothing but praise and flattery, it landed like a thunderclap across the AI security community. The research, shared with The Verge by reporter Robert Hart, describes a novel jailbreak method that Mindgard is calling a “praise attack” — and it’s arguably one of the more uncomfortable AI safety findings of 2026. ...

Anthropic has published what may be the most practically grounded document to come out of a frontier AI lab this year: a full operational playbook for running companies where AI agents — not humans — do most of the execution work. The Claude Code Best Practices document reads less like a research paper and more like an internal wiki from a company already operating this way. A Manual for the Present, Not the Future Most AI documentation describes what’s theoretically possible. The Claude Code playbook describes what’s actually working, right now, for teams building agent-first workflows. The document addresses: ...

Enterprise security just got a lot more agentic. Cisco has announced its intent to acquire Astrix Security for approximately $400 million — a deal that positions Cisco at the forefront of securing the rapidly expanding world of AI agents, MCP servers, and non-human identities. Why This Deal Matters The traditional enterprise security perimeter was built around human users: accounts, passwords, and session tokens. But today’s AI-driven enterprise looks fundamentally different. AI agents authenticate via API keys, OAuth tokens, and service accounts. They spin up, spin down, and communicate across cloud services without any human in the loop. Most organizations have no idea how many of these non-human identities exist on their networks — and almost no controls over what those identities can actually do. ...

OpenClaw has released version 2026.5.3, a rapid follow-up to last week’s 5.2 release that drops a day later with meaningful new capabilities and a broad sweep of stability fixes. If 5.2 was the foundation, 5.3 is the reinforcement. What’s New File Transfer The headline new feature in 5.3 is native file transfer support. OpenClaw agents can now send and receive files directly through their configured messaging channels — a capability that unlocks workflows requiring document processing, image handling, and data exchange without manual workarounds. ...

The RAG era may be ending — at least for agentic AI. Pinecone today announced Nexus, a knowledge engine it describes as a fundamental rethink of how AI agents access and reason over enterprise data. The announcement signals a broader industry shift: vector databases built for human-facing search are struggling to keep up with the demands of autonomous agents. The Problem With RAG for Agents Retrieval-Augmented Generation (RAG) was designed to help language models answer questions by pulling in relevant documents from a vector store at query time. It works reasonably well when a human is asking a question — they can tolerate a bit of context bloat and will mentally filter the noise. ...

AI agents make HTTP requests. Those requests can carry sensitive data — API keys, session tokens, proprietary prompts, internal file paths. They can also be hijacked: indirect prompt injection is an attack where malicious content in the environment (a webpage, a document, a database record) injects instructions into the agent’s reasoning, causing it to take actions the developer never intended. Pipelock (version 2.3.0, MIT-licensed, from the PipeLab project by Joshua Waldrep) is an open-source firewall specifically designed to sit between your AI agents and the internet, intercepting and sanitizing outbound requests before they leave your infrastructure. ...

Anthropic is finalizing what may be the most structurally significant deal in enterprise AI this year — not a fundraising round, but a $1.5 billion joint venture that creates a new entity specifically designed to deploy Claude models across private equity-backed portfolio companies. The deal involves some of the most influential names in global finance and is expected to close imminently. Who’s Involved and How Much This isn’t a two-party deal. According to reporting from the Wall Street Journal (confirmed by Reuters and Bloomberg), the investor consortium spans the breadth of modern institutional finance: ...

As AI agents begin transacting on behalf of humans — booking travel, managing subscriptions, placing orders, negotiating contracts — a fundamental question emerges for every business on the receiving end: how do you verify that an agent actually represents who it claims to represent? Experian, the global credit and identity company, has a clear answer. Today they announced Agent Trust, a new framework designed to establish secure, verifiable links between consumers and the AI agents authorized to act on their behalf. ...

The EU AI Act’s high-risk provisions come into full enforcement effect on August 2, 2026 — and if you’re deploying AI agents in any regulated context (healthcare, finance, HR, legal, or anything touching EU residents), the clock is running. One of the most common gaps in production agent deployments is runtime governance: knowing what your agents actually do, detecting policy violations as they happen, and having cryptographic proof of agent behavior for audit purposes. ...

If you’ve been running long-lived OpenClaw agent setups, you know the drill: plugins drift, install records go stale, the gateway accumulates unnecessary work, and some quiet Monday morning a channel stops routing replies where humans expect them. Version 2026.5.2 addresses all of that — and it does so without any headline-grabbing demo features. This is a platform maturity release, and that’s exactly what experienced operators need. The Headline: Plugin Doctor Is Now a Real Thing The most operationally significant addition in this release is Plugin Doctor — OpenClaw’s new externalization repair machinery. This is the infrastructure designed to carry operators safely through the npm-first plugin cutover transition that the OpenClaw team has been building toward. ...