AI Safety news and analysis on Web Pulse: 12930 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
KU Leuven's Centre for IT & IP Law (CiTiP) will host a conference on 10 December 2026 in Leuven to examine the EU's Digital Omnibus proposal and its impact on democracy, digital sovereignty, AI fairness, and other areas.…
PolicyLayer argues that while OAuth in the Model Context Protocol (MCP) improves on static API keys, it introduces token-management challenges at scale. The company proposes an MCP gateway that centralizes OAuth flows, t…
A developer proposes an MCP gateway as a missing control layer for AI agent fleets, addressing the Model Context Protocol's lack of permission models. The gateway sits between AI clients and MCP servers, enforcing tool f…
A developer at PolicyLayer describes a solution to prevent AI agents from posting to sensitive Slack channels like #general. The approach uses channel allowlists and denylists in an MCP server policy, rather than rate li…
A critical authentication bypass vulnerability (CVE-2026-48558) in SimpleHelp's remote monitoring and management tool allows unauthenticated attackers to create a Technician account and gain full access to managed endpoi…
Two teenagers were killed and two others hospitalized after a two-vehicle crash on Highway 152 in Santa Cruz County on Sunday evening. The collision involved a 2011 Infiniti G37 and a 2026 Toyota Camry, with the Infiniti…
Anthropic's Claude Fable 5 launched on June 9, 2026, and was so capable that the US government issued an export control directive within 76 hours, banning its export due to national security concerns.
X (formerly Twitter) released an official MCP server, xmcp, that exposes 131 tools from the X API v2 to any connected agent with no access control, permissions model, or rate limiting. The tools include destructive actio…
Probably, a San Francisco-based AI startup founded by Peter Elias, raised $9 million in seed funding led by Andreessen Horowitz to build a reliability layer that prevents AI hallucinations. The company targets 99.99% acc…
Databricks announced new innovations for Unity AI Gateway at Data + AI Summit 2026, extending governance to AI assets including models, agents, MCP services, and skills. The platform adds unified cost management, runtime…
Databricks announced new Unity Catalog features at Data + AI Summit 2026, including Unity AI Gateway for runtime governance of AI agents, Glossary and Domains for business context, and cross-cloud/region addressability. …
The US Justice Department invoked national security to defend xAI's unpermitted gas turbines, arguing that shutting them down would harm AI innovation supporting military operations. The NAACP sued over emissions from th…
A developer warns that the PostgreSQL MCP server exposes a single 'query' tool that executes raw SQL, allowing AI agents to run destructive commands like DROP TABLE or DELETE without a WHERE clause, despite being describ…
The official GitHub MCP server exposes 83 tools, including destructive operations like file deletion and repository creation, without any permission model. PolicyLayer's Intercept proxy enforces deterministic YAML polici…
A developer warns that coding agents using the popular filesystem MCP server can delete or overwrite files without restrictions, as the server's 12 tools have no rate limits or confirmation steps. The developer introduce…
A developer warns that AI coding assistants with access to Docker MCP servers can inadvertently delete all containers, images, and volumes on a machine. The ckreiling/mcp-server-docker exposes 19 tools, including destruc…
Cloudflare's official MCP server gives AI agents access to 30 tools including DNS record deletion, which can cause production outages. A developer introduced Intercept, a policy engine that sits between agents and the MC…
Apple published a support document detailing its Terminal anti-scam warning, introduced in macOS Tahoe 26.4, which blocks pasted commands that could be harmful. The alert warns users about potential malware from scammers…
Brent Ozar is discontinuing SQL ConstantCare and the Consultant Toolkit due to increasing AI-driven security risks, lower-than-expected cloud adoption, and insufficient product adoption. The company is also laying off lo…
Tailscale launched new capabilities for its Aperture AI access and control platform to help IT teams manage shadow AI, providing a stable layer for identity, access, and control across evolving models, tools, and data so…