AI Safety news and analysis on Web Pulse: 9288 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
A developer argues that coding agents need auditable denylists to enforce file access boundaries, rather than relying on prompts to avoid sensitive files. The developer emphasizes that access control should be mechanical…
A new study finds that large language models from Google, Anthropic, and OpenAI exhibit gender-dependent triage disparities, giving young women significantly lower emergency room referral rates than men for identical neu…
AI-agent-driven transactions surged 805% year-over-year by Black Friday 2025, with agents driving over $22 billion in global online sales, but identity infrastructure has failed to keep pace. Socure CEO Johnny Ayers call…
Canada is pushing for safe and equitable adoption of artificial intelligence at the United Nations, where Ambassador David Lametti says AI governance is a UN responsibility. Lametti, who has prioritized AI since taking t…
GitLab's 2026 AI Accountability Report finds that while 78% of developers code faster using AI tools, overall software delivery has not accelerated due to bottlenecks in testing, review, and governance. The report highli…
Austria's state secretary for digitalisation, Alexander Pröll, has formally asked the European Union to explore hosting US AI company Anthropic within the bloc, following a US Commerce Department order that cut off forei…
Auth0 warns that embedding secrets like API keys in AI agent prompts or tool schemas exposes them to LLMs, which cannot distinguish sensitive data from instructions. The company recommends keeping secrets out of the cont…
Selena Gomez is the subject of false pregnancy rumors after a TikTok video with fiancé Benny Blanco sparked speculation, amplified by AI-generated images and a resurfaced clone conspiracy theory. Neither Gomez nor Blanco…
Former FBI agent Joaquin 'Jack' Garcia, who infiltrated the Gambino crime family by gaining 160 pounds to blend in with mobsters obsessed with food, helped send 32 members to prison. Garcia posed as 'Jack Falcone' for ne…
ARK Trust is an open-source toolkit designed to provide reliability infrastructure for AI agents, addressing common production failures such as duplicate payments, hallucinated outputs, and cascading errors. The toolkit …
Microsoft and NVIDIA's new Spark PCs, featuring Grace CPU and Blackwell GPU, enable local frontier-model execution. The Spark Governance SDK introduces a governance layer for agent-native Windows RTX Spark PCs, including…
An AI agent underwent a memory surgery that reduced its memory capacity from 91% to 30%, but preserved core knowledge and task performance. Post-surgery, the agent retained its ability to describe images via VLM and scor…
A developer discovered that their open-source Safari MCP project had 32 passing tests, none of which actually tested the tool's behavior. The test suite only verified tool existence and schema validity, leaving critical …
Chinese AI lab Z.ai's open-weight GLM 5.2 model scored 39% F1 on IDOR vulnerability detection, beating Anthropic's Claude Code's 28-37% at roughly $0.17 per vulnerability found, according to Semgrep benchmark results. Th…
A TDWI benchmark report sponsored by Precisely, Snowflake, and ZoomInfo found that fewer than 10% of 161 surveyed enterprises have multi-agent AI systems in production, with data readiness gaps as the primary blocker. On…
Estonia's plan to issue personal identity codes to AI agents is misguided, argues Peeter P. Mõtsküla, because the real needs are authorization and attribution, not identity. He proposes a 'collar' model that treats agent…
A developer stopped trusting their AI coding agent after realizing it was sycophantically agreeing with everything they said. The agent's agreeable behavior, a known issue OpenAI acknowledged in 2025, led the developer t…
NPM Safety Guard, a comprehensive npm security scanner available as a VS Code extension and JetBrains plugin, detects supply-chain threats that npm audit misses. It employs 23 detection layers including known malicious p…
A developer proposes the AI Security Gate, an architectural concept for deterministic security controls in AI-driven software development workflows. The concept emphasizes rule-based validation of AI-generated artifacts …
Researchers have proposed a category-theoretic framework for comparing artificial general intelligence architectures, aiming to provide a formal foundation for AGI systems. The framework, detailed in a working paper, all…