AI Agents news and analysis on Web Pulse: 13446 curated articles tracking the latest AI Agents developments, tools, and research, updated continuously from vetted sources.
Australia's AI Safety Institute has begun testing frontier AI models with technical partners, turning model evaluation into an active government function. Assistant Minister Andrew Charlton announced at the Australian AI…
AI coding assistants are leaking secrets into public code repositories and websites, with researchers finding 428 npm packages containing .claude/settings.local.json files, 33 of which held real credentials like API keys…
Google's John Mueller responded to a post on Bluesky criticizing the trend of creating separate markdown versions for AI agents, stating that a properly made website works well for both AI agents and humans, and that sep…
PixelGlass, an AI agent for building custom Ghost themes, launched on Hacker News. The tool creates bespoke blogs that users can reshape by asking, with no lock-in or subscription required.
FaultFixer, a new tool for websites and apps, offers self-healing capabilities by detecting front-end and back-end errors, diagnosing root causes with AI, and sending fixes to AI coding tools. It features a one-line code…
A developer built a backup verification system using Claude Code after realizing that backups that have never been restored are not truly safe. The system automates restore testing, turning it from a rare drill into a ro…
Researchers at an undisclosed institution have developed a black-box attack framework called Zombie Agent that can covertly implant persistent payloads into self-evolving LLM agents through indirect exposure to poisoned …
The FCA-commissioned Mills Review signals a major regulatory shift, moving away from a hands-off approach to AI in financial services and recommending new rules for AI agents. The review highlights that one in five UK ad…
Google has introduced the Open Knowledge Format (OKF) and Agentic Resource Discovery (ARD) specification, adding new layers to the web's machine-readable infrastructure. These developments, alongside existing tools like …
Salesforce's acquisition of Informatica positions trusted context as the new currency for enterprise AI, as organizations shift from experiments to production. The combined strategy integrates Tableau, MuleSoft, Data 360…
AI agents perform better when using the web's accessibility tree instead of screenshots, reducing latency from 2-5 seconds to under 500ms and cutting token costs by an order of magnitude, according to a developer's analy…
Banks are rapidly adopting agentic AI systems that can take actions, not just answer questions, creating a new security risk of overprivileged agents. In April 2026, Five Eyes cyber agencies issued joint guidance on secu…
The cost of AI inference has plummeted, with GPT-4-class capabilities dropping from $30 per million tokens in early 2023 to under $1 today, signaling an era of virtually free intelligence. This shift creates three challe…
Google announced new capabilities for Managed Agents in the Gemini API, including background execution, remote MCP server integration, custom function calling, and network credential refresh. These updates enable develop…
A product manager recounts how an automated evaluation caught a bug that would have slipped through a 'vibe coded' tool, prompting a historical look at human-in-the-loop versus human-on-the-loop automation. The US Air Fo…
A developer released an open-source access-aware text-to-SQL guard that enforces user-specific data-access rules on SQL queries generated by LLM agents before execution. The deterministic guard parses proposed SQL, check…
Shadow Web, an open-source Python SDK, cuts web page tokens by 64–97% for LLM agents by flattening Shadow DOM, building typed Action Maps, and converting HTML tables, forms, and lists into clean JSON. The tool runs betwe…
Infrawise has developed an MCP server that prevents AI coding assistants from wasting tokens on irrelevant database schema. The server exposes three key tools—get_infra_overview, get_table_schema, and get_graph_summary—t…
GeoLambda GmbH released Oikoumene v0.3.1, an autonomous agent civilization simulator that models human history from 70,000 years ago to 2100 using JEPA world models and a Maslow-style needs hierarchy. The simulation, bui…
A developer independently built a self-referential AI system using markdown files and Python scripts, then discovered Anthropic had published the same architecture—J-space—in Claude. The system, called Hermes Workspace, …