AI Ethics news and analysis on Web Pulse: 16845 curated articles tracking the latest AI Ethics developments, tools, and research, updated continuously from vetted sources.
About half of all companies that laid off workers and replaced them with AI are now rehiring at greater expense, according to CNBC. Klarna, which replaced 700 customer service agents with an AI chatbot, saw customer sati…
Organizations are deploying AI as an autonomous actor that makes decisions leaders once owned, eroding human judgment, according to professor Carly and organizational advisor Jenny. They cite the case of "Elena," a chief…
AI-generated 'virtual staging' images are flooding UK property listings, with 52% of estate agents planning to adopt AI tools by 2026, according to research from Alto Software. Critics argue the deceptive images mislead …
A developer uses ChatBrat AI, a chatbot app, to simulate conversations with their deceased grandparents, who died in a car accident. The app generates responses based on memories and photos the user uploads, allowing the…
Researchers at Michigan Technological University demonstrated an AI overthinking attack that can immobilize a robot for over a minute by exploiting large vision-language models that process text appearing in camera frame…
EG-VAR, a Lean 4-based architecture for AI reasoning, achieved a flawless 120 out of 120 on TableBench numerical reasoning tasks and maintained 100% source fidelity during counterfactual stress tests, outperforming same-…
New research reveals that up to 77% of features in degraded Sparse Autoencoders (SAEs) are functionally inert, and even in well-trained models 9% of matched features are dead, challenging the interpretability claims of S…
Twenty-six Meta employees have sued the company, alleging it used artificial intelligence tools to tag workers for mass layoffs after they took protected leave, including maternity leave and disability accommodation. The…
Job hunters are using AI tools to cheat during interviews, but the same candidates often underperform once hired, according to a Bloomberg report. The trend highlights a growing disconnect between AI-assisted hiring tact…
A recent analysis of Chronic Kidney Disease prediction studies found that high-leakage studies reported accuracy rates up to 95.48%, while studies without data leakage achieved around 80.2%, revealing a 15% gap due to me…
Support Vector Attention (SV-Attention), a new approach to memory in machine learning, achieves certified token deletion with a rare-item recall of 0.86 compared to 0.32 for other methods, according to researchers. The t…
Google's new AI-generated search summaries, which provide direct answers instead of links to explore, are undermining users' curiosity and depth of understanding, according to writer Alan Noble. The shift removes the ser…
A new clinical AI system removes therapeutic decision-making from the large language model (LLM), using a deterministic pipeline that scores client state, selects an action from an admission table of nine therapeutic sch…
A large-scale empirical study using the AIDev dataset found that 38.9% of autonomous coding agent-generated pull requests contain at least one security smell, with supply chain integrity issues accounting for 82.3% of al…
Publisher ad supply fell by up to 40% in Q2 2026 as AI-driven zero-click search reduced traffic to news sites, according to benchmarking data from Ozone shared exclusively with Digiday. The data, tracking approximately 2…
Reddit's evp ads monetization Roelof van Zwol admitted at Cannes that the company is "intensely debating" whether selling its community data to train OpenAI's ChatGPT and Google's Gemini undercuts the data advantage its …
Check Point's AI Security Report 2026 documented intrusions over the past twelve months in which AI ran exploitation workflows autonomously, generating thousands of commands across dozens of sessions with minimal human d…
A new arXiv paper argues that the alignment techniques used to improve large language models like GPT-2 represent an expression of 'optimization culture' that conflates measurable improvement with value, and that this op…
A systematic review of 19 machine learning studies on early Chronic Kidney Disease (CKD) prediction found that high data leakage inflates reported accuracy to an average of 95.48%, compared to 80.2% for leakage-free stud…
GRID (Grammar-Railed Decoding), a grammar-constrained decoding engine for enterprise SQL generation, achieves a median per-token mask latency of 3.6-6.7 microseconds using Rust kernels, outperforming llguidance at p50 an…