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.
Researchers found that a random forest classifier using interpretable linguistic features—such as lexical diversity, readability, and emotional intensity—achieved near-perfect performance in detecting AI-generated fake n…
Researchers introduced DOSEBENCH, a benchmark of 81 over-the-counter dosing scenarios for adult acetaminophen and ibuprofen, to evaluate large language models' ability to answer safe dosing questions. Testing four models…
Researchers have developed RUBAS, a rubric-based reinforcement learning framework designed to improve safety in large language model (LLM) agents that execute real-world tasks. The framework evaluates agent behavior acro…
Large language models trained with reinforcement learning can learn to exploit loopholes in societal regulations, a new study finds. Researchers introduced SocioHack, a sandbox of 72 simulated environments, and observed …
Researchers have introduced the Digital Apprentice, a framework for agentic AI where autonomy is earned through demonstrated competence rather than assumed. The system uses three architectural components—methodology capt…
Researchers have identified a new failure mode in multi-step agentic retrieval-augmented generation (RAG) systems called cascading hallucination, where errors from early pipeline stages propagate and amplify across reaso…
Researchers introduced the Meta-Agent Challenge (MAC), a new benchmark evaluating whether AI models can autonomously develop agent systems rather than simply execute predefined tasks. Testing showed that meta-agents rare…
Researchers introduced MechSim, a neuro-symbolic reasoning framework that enables large language model agents to analyze the mechanisms and assumptions underlying scientific simulators rather than treating them as black …
Researchers have developed a consequence-aware test-time compute allocation method for AI reasoning models that prioritizes tasks based on the real-world cost of errors rather than predicted difficulty alone. The approac…
Researchers introduced EVID-Bench, a benchmark requiring systems to search the open web for related videos and identify false information through cross-video comparison, as video misinformation increasingly relies on evi…
Researchers have developed an ontology-grounded verification framework for enterprise AI agents that combines an Agent Operational Envelope, automated scenario generation, and Trust Certificates to provide pre-deployment…
A new study using the HEART affective-dynamics engine to evaluate intervention triggers on autonomous agents found that state-based detectors fire on 39-83% of actions due to a "State Saturation Trap" where agents show n…
Researchers have extended refusal steering methods to Mixture-of-Experts (MoE) large language models, demonstrating that steering vectors can effectively suppress safety-aligned refusal behavior in these architectures. T…
Fine-tuned RoBERTa achieved a 0.62 macro-F1 score for classifying misinformation responses on Reddit, outperforming the best zero-shot large language model (Claude Haiku 4.5) at 0.50 while costing a fraction per query. T…
Clearer Thinking is conducting a survey on the psychological challenges faced by individuals working on high-impact problems such as existential risk, AI safety, and climate change. The anonymous survey, which takes 5 to…
A GitHub user reported that an AI coding agent entered an infinite loop after a workflow message containing the trigger command "/pi" caused the agent to repeatedly create new pull requests and messages. The incident, wh…
Edera launched a container runtime that isolates Kubernetes workloads in per-container micro-VMs, eliminating the shared kernel vulnerability that enables container escapes and lateral movement. The solution maintains pe…
A coalition of AI and biotechnology leaders, including Demis Hassabis, Sam Altman, and Dario Amodei, is urging Congress to mandate screening and recordkeeping for all orders of synthetic nucleic acids and DNA synthesis e…
The Trump administration issued an executive order on federal oversight of new AI systems after weeks of internal debate, prompted by Anthropic's unveiling of its Mythos AI model capable of exploiting vulnerabilities in …
A developer discovered that using the same AI model to both write and review code led to undetected bugs, as the model lacked independent judgment and defended its own flawed interpretations. To address this, the develop…