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.
OpenAI CEO Sam Altman and Anthropic CEO Dario Amodei walked back their earlier predictions that AI would eliminate millions of jobs, acknowledging that automation is instead creating more work. The shift reflects Jevons …
Organizations deploying shared AI agents face data exposure risks when agent memory lacks access controls, as one agent's stored information — such as private commission details or confidential negotiations — becomes acc…
European enterprises purchasing AI platforms in 2026 must distinguish between data residency and data sovereignty, as the latter requires control over infrastructure, legal regimes, model routing, and audit trails across…
Regulated enterprises are adopting AI decision receipts—structured records of agent workflows—to provide compliance, audit, and security teams with a verifiable account of what an AI agent did, which data it used, and wh…
Enterprise AI agent platforms require governance, private RAG, model routing, observability, tool access control, and evaluation to move beyond demo capabilities into production-ready systems for regulated industries. Ve…
The EU AI Act requires enterprises deploying on-premises AI systems to compile an evidence pack before production, documenting system identity, risk classification, data sources, model details, and human oversight contro…
Kadhir Mani argues that Conflict-free Replicated Data Types (CRDTs) are insufficient for collaborative editing when AI agents are involved, because CRDTs guarantee convergence but not coordination. The author describes h…
OpenAI's new ChatGPT memory feature shifts from storing isolated facts to maintaining a continuous, evolving synthesis, raising questions about machine consciousness. A new AAAI paper by Scott Hughes and Karen Nguyen pro…
A developer built a system that automatically processes in-app user feedback by sanitizing submissions, classifying them, and dispatching eligible bug reports and feature requests to a coding agent that opens pull reques…
AI enthusiasts face an existential threat from competitors who adopt rapid AI integration, while skeptics face an existential threat from reliability degradation and institutional knowledge loss caused by shipping code f…
ABC's Four Corners, reported by Steve Cannane, investigates the global race to develop more powerful AI systems and reports that the world's most powerful AI companies seek to invest billions in Australia. The program, a…
The industry is transforming Non-Human Identity (NHI) to reduce attack surfaces by replacing vulnerable credentials like passwords and API keys with public-private key pairs and rapid rotation. Emerging standards from th…
Anthropic called for a slowdown in the rapid development of artificial intelligence, stating that a more measured pace "would be good for the world." The plea for caution came the same week the AI company beat its archri…
South Korea will require all internet forums and community websites to scan every user-uploaded image and video using AI censorship tools starting July 1st, under revised telecommunications laws. Website owners must purc…
Researchers have developed a post-training procedure called Latent Agents that distills multi-agent debate into a single large language model, achieving comparable or superior reasoning performance while using up to 93% …
Anthropic co-founder Jack Clark told BBC Newsnight that the AI industry needs a way to slow progress, warning systems could reach a point where they "develop without human input." Clark said Anthropic's chatbot Claude is…
A Hacker News user posted an open thread asking tech workers how they are feeling, citing a sense of collective darkness. Commenters responded with advice on maintaining a positive outlook through action and questions ab…
The Electronic Frontier Foundation's Senior Policy Analyst Dr. Matthew Guariglia testified before the House Homeland Security Subcommittee on Cybersecurity and Infrastructure Protection, warning that government use of ge…
The Pink cybercriminal group has adopted a new tactic of posing as helpdesk support to trick victims into handing over credentials. The group calls targets claiming to be from IT support and directs them to fake login pa…
A developer detailed the technical challenges of deploying secure Retrieval-Augmented Generation (RAG) and autonomous AI agents in healthcare, highlighting risks like prompt injection, semantic collision, and context ove…