AI Safety news and analysis on Web Pulse: 13553 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
A final-year mathematics and computer science undergraduate is questioning whether pursuing a career in theoretical robotics, specifically in continual learning for human-like robot adaptation, could inadvertently accele…
AI agents can fail silently without triggering errors, wasting resources by repeating the same broken commands hundreds of times. A developer built AgentSonar, a tool that installs hooks into Claude Code to monitor tool …
Two lawyers were sanctioned by the Ninth Circuit on June 3 for relying on generative artificial intelligence to draft a legal brief, resulting in numerous errors and false statements to the court. The court ordered each …
President Trump on June 5 signed a National Security Presidential Memorandum directing the U.S. national security enterprise to accelerate AI adoption, including rapid onboarding of advanced AI models and construction of…
A small study testing whether wrapping untrusted prompt inputs in mock tool calls could improve language model robustness found the technique did not broadly help across three LLM-as-a-Judge tasks and in some cases made …
A developer has outlined the persistent technical limitations of large language models (LLMs), including hallucination, static knowledge cutoffs, statelessness, limited context windows, and compounding errors in multi-st…
Anthropic proposed that leading AI labs develop a system to potentially pause frontier AI development, comparing the need for oversight to nuclear weapons monitoring. Former Senator Mitt Romney endorsed the idea, calling…
A developer's analysis of AI-generated code in the rsync utility found that Claude introduced subtle but significant bugs, raising concerns about AI reliability in critical codebases. Separately, researchers traced a pow…
Suhana Sayyad, an MSc Cybersecurity student at TUS Athlone, built a free, open-source interactive tool that maps 30 EU AI Act obligations to their equivalents in NIST AI RMF and ISO 42001. The tool provides strength rati…
A developer built gvnr, an open-source MCP server that provides per-agent spend caps and a human approval gate for AI agents. The tool hard-stops agent calls when budgets are exhausted and sends mobile approval links for…
Researchers introduced two new consistency training methods, AttCT and MLPCT, which enforce output consistency in attention weights and MLP post-activations respectively, and applied all four existing and new methods to …
A developer built a Puppeteer MCP server designed to enforce safety controls for AI-powered browser automation. The tool allows agents to click, fill forms, run JavaScript, and capture pages, but restricts these actions …
The June 5, 2026 episode of the Ctrl-Alt-Speech podcast, "Generous To A Default," covered Meta's global rollout of new content settings for teen accounts, a Krebs on Security report detailing attackers abusing Meta's AI …
The National Security Agency is using Anthropic's Mythos AI model for offensive cyber operations likely targeting China and Iran, with the company embedding engineers at the agency to customize the technology. The collab…
A university undergraduate student, referred to as "Joe," spent eight weeks building an AI agent that he claimed could automatically verify the MTQQ protocol specification, presenting the completed project to his profess…
Prime Minister Mark Carney launched Canada's AI for All strategy on June 4, directing more than $200 million CAD to the country's three national AI institutes and the Canadian AI Safety Institute. The funding supports 70…
Cloudflare CEO Matthew Prince reported that automated traffic now accounts for 57.5 percent of HTTP requests on the company's network, surpassing human traffic for the first time in internet history. The milestone arrive…
India's electronics and IT secretary S Krishnan said the country needs greater "strategic autonomy" and control over hardware underpinning AI and biometric systems, citing supply-chain and industrial-espionage risks. Spe…
The US National Security Agency is using Anthropic's Claude Mythos model for offensive cyber operations, with roughly half a dozen Anthropic engineers embedded inside the agency, according to the Financial Times. Axios r…
Meta has embedded facial recognition code, internally called "NameTag," into the Meta AI smartphone app used by its smart glasses, according to investigations by Wired and the Electronic Frontier Foundation confirmed Thu…