AI Safety news and analysis on Web Pulse: 10311 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
The US and Israel launched massive strikes on Iran on February 28, 2026, claiming an imminent nuclear threat, after decades of sabotage and brinkmanship over Iran's nuclear program. The film traces the shadow war from th…
A developer building the NENE2 PHP framework argues that Model Context Protocol (MCP) tools should call application boundaries, not production databases directly. The framework enforces this by mapping MCP tools to docum…
Two AI agent security startups, Arcade.dev and Convey, raised a combined $98 million in 72 hours, signaling a market gap as 80.9% of technical teams have AI agents in production but only 14.4% have full security approval…
Armorer Labs, the team behind Armorer and Armorer Guard, argues that AI agents should run with their own session identity rather than inheriting the human user's credentials. The company identifies three failure modes—in…
Varonis Threat Labs researchers Doron Kapah and Mark Vaitsman launched Breach at the Beach, a Capture The Flag (CTF) training experience designed to teach security practitioners how to detect and respond to data exfiltra…
The Trump administration's chief digital and artificial intelligence officer argued that shutting down xAI's Colossus 2 data center in Memphis would threaten national security, as the facility powers Grok AI used by the …
A medical AI system that reports 99% confidence in a pneumonia diagnosis may be correct only 80% of the time, highlighting a calibration problem where confidence scores do not reflect actual accuracy. Research shows mode…
Upwind Security launched AI Sensor for Endpoints, extending its cloud and AI security platform to developer workstations to provide visibility into AI-related activity across endpoints, cloud, and SaaS platforms. The rel…
360 Security Technology founder Zhou Hongyi warned that China must develop its own version of Anthropic's Mythos AI model to counter cybersecurity threats, calling the US technology a 'cyber nuclear weapon' that has demo…
Vicfuse introduced a UL Class fuse series designed for modern AI infrastructure and industrial protection, offering application-driven categories such as Class CC, J, G, and T to address varying requirements for interrup…
Rapid7 VP of Global Government Affairs and Public Policy Sabeen Malik says frontier AI is forcing security leaders to adopt a new operating model that integrates vulnerability discovery, compliance, and operational resil…
OpenAI and Azure OpenAI's JSON mode endpoints cannot correctly emit escaped Unicode characters like é, instead producing NUL bytes that break systems such as PostgreSQL. The constraint is undocumented and affects any UTF…
Apple raised Mac mini prices from $599 to $799 and increased iPad Pro and iPad Air prices, citing memory costs driven by AI infrastructure demand. The shift in DRAM supply toward high-bandwidth memory for AI servers is s…
A new paper by Simon Willison, highlighted on Bruce Schneier's blog, reveals that large language models suffer from role confusion, enabling prompt injection attacks. The research shows that models learn to recognize tex…
Canadian podcaster Dr. Jordan Peterson announced he will release recorded lectures on Sundays while battling akathisia, a neurological disorder caused by a depressant medication. The 64-year-old author and provocateur sa…
China claims to have developed an AI 'cyber nuclear weapon' that matches the capabilities of the fictional Mythos system, raising concerns about a new era of cyber warfare.
NakshGuard, an open-source on-premises reverse proxy, detects and blocks runaway loops in AI agent traffic before they consume excessive API tokens. It sits between agents and LLM APIs like OpenAI and Anthropic, applying…
RunZero announced runZero 5.0, a platform update that unifies the exposure management lifecycle into an automated workflow to help organizations defend against AI-fueled threats. The release enables security teams to dis…
Anthropic's Mythos AI security tool, touted as too dangerous to release, found only one low-severity vulnerability in the heavily audited curl codebase after a privileged deployment. The curl project lead dismissed the h…
The European Union is set to join Pax Silica, a US-led initiative to coordinate AI chip supply chains and export controls against China, despite France's criticism that the pact amounts to colonization and contradicts th…