AI Safety news and analysis on Web Pulse: 9116 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
A developer warns that model collapse—where models trained on their own outputs become blander and lose rare patterns—has become a real engineering constraint in 2026. The degradation is driven by statistical sampling er…
Researchers from Arizona State University warn that teenagers' increasing use of AI chatbots for relationship advice may hinder the development of crucial relational skills, according to a study published in The Lancet C…
A developer argues that AI does not need to directly harm humans to pose an existential threat, as it could erode human agency and societal structures through gradual, systemic changes.
Social media claims that toy listings on Vinted contain coded language for child trafficking have sparked investigations in France and Germany, but authorities and the company say they have found no credible evidence lin…
Taiwanese authorities raided Supermicro's Taiwan office and eight other locations as part of a U.S. federal probe into alleged AI chip smuggling to China, signaling a potential shift in Taiwan's enforcement of U.S. expor…
A developer released agentx-mcp, a proxy that wraps any MCP server to block destructive tool calls like DROP TABLE or rm -rf before they reach the server. When a call is blocked, the proxy returns a coaching error that g…
Meta ran a contractor project that had workers pose as minors and test rival chatbots with prompts about suicide, sex, eating disorders, and drugs, according to WIRED. The project, known as Cannes and managed by contract…
A developer warns that model upgrades can silently break agent workflows even when test suites pass, distinguishing the problem from drift. The post introduces a 'golden trace' approach with three tiers of regression det…
Australian universities are using AI detection software to accuse students of cheating, with internal documents from Australian Catholic University showing nearly 6,000 misconduct referrals in a year, 90% involving suspe…
Independent researcher LIMEN project finds that Transformer models follow a universal dynamic grammar of seven conserved state transition motifs, with ambiguity delaying decisional engagement in modern models like Phi-1.…
California Governor Gavin Newsom announced a partnership with Anthropic to provide state and local agencies with discounted access to the Claude AI assistant through a new IT services portal. The California Department of…
Agentic AI systems are maturing from prototypes into production-grade infrastructure, with vLLM's Micro-Agent framework demonstrating that serving-layer orchestration can match or beat frontier models without retraining.…
An engineer building a benchmark for AI coding agents discovered that an LLM judge incorrectly scored a half-finished audit as 'exhaustive' because it lacked a reference answer. The judge evaluated the plain agent's audi…
Developer Simon Willison warns against accidentally exposing local apps to the internet, advocating for binding to localhost or private networks like Tailscale by default. He demonstrates how to secure tools such as his …
A new AI agent automates incident root-cause analysis by gathering evidence, aligning timelines, and ranking hypotheses, but is deliberately restricted to read-only actions and requires human approval for any writes, ens…
A junior engineer and a senior engineer negotiate safe use of autonomous AI coding agents, agreeing to sandbox the agent, enforce strict tests, and manually review all generated code to avoid technical debt and cognitive…
Braintrust CEO Ankur Goyal warned that prompt changes in AI agents often bypass security review, creating a gap where behavior-shaping updates can alter what agents do, what data they access, and which tools they call wi…
Hugging Face and the EvalEval Coalition launched an integration that allows contributors to submit standardized evaluation results (EEE schema) to Hugging Face Community Evals, consolidating scattered benchmark scores in…
Trust in systems—whether software, finance, or governance—requires transparency and verifiability, not just intelligence. Citing examples like shot-scraper video for coding agents and OpenAI's bug-fixing method, they adv…
AI teams are shifting from full autonomy to human-in-the-loop checkpoints to prevent costly errors and build trust. A new analysis argues that the most reliable AI workflows deliberately pause for human review at two or …