AI Ethics news and analysis on Web Pulse: 17044 curated articles tracking the latest AI Ethics developments, tools, and research, updated continuously from vetted sources.
Two jury verdicts against Meta and Google in late March 2026 signal emerging liability risk for software platforms designed to drive prolonged engagement, with implications for insurers as design-centered theories may ex…
Anthropic refunded disputed Claude Code charges this month after auditing startup Vaudit found about $1.7 million in mistaken overcharges across 60 companies, with Amazon, Google, Microsoft, Anthropic and OpenAI together…
China is rapidly deploying AI across sectors, from medical avatars to food-delivery drones on the Great Wall, while expanding state surveillance with facial recognition and data tracking. The government's embrace of AI r…
Four years into the AI revolution, the idea that AI makes it possible to build everything in-house and kill SaaS is dangerously misleading, argues a seasoned software developer. A CTO who replaced Stripe with Claude AI c…
Palm Beach International Airport has been renamed in honor of President Donald Trump, but the new metallic gold logo featuring a heavily altered Great Seal eagle is being widely criticized as AI-generated due to glaring …
Meta pulled its new AI image tool, Muse Image, just 72 hours after its July 8 launch due to backlash over privacy concerns, with the feature automatically altering public images without explicit consent. The company, thr…
A developer building an AI content detector discovered that high accuracy claims are misleading, as even a 0.1% false positive rate can wrongly accuse thousands of real people. The tool's probabilistic nature and lack of…
C1's founder and CEO, publishing a six-month dogfooding experiment with 125 employees across engineering, GTM, security, and support, argues that winning the agentic shift depends on cultural adaptation and security-enab…
ConceptSMILE, a model-agnostic auditing framework, assesses the reliability of concept-based AI explanations by extending SMILE's perturbation-based logic. In a case study on retinal fundus images, MedSAM showed superior…
Urban mining's future depends on AI that supports human auditors through defensible, not just accurate, pre-demolition assessments, according to a new analysis. The integration of explainable AI and domain knowledge grap…
Researchers propose a semantic framework to assess AI output correctness by distinguishing between domain knowledge, reference sources, and system capabilities, aiming to improve reliability and trust. The framework defi…
LongMedBench, a new benchmark for long-term clinical decision-making using real-world EHR data from 335 patients with an average of 19.72 inpatient visits and 44.91 medical events per visit, reveals that current LLMs str…
Samsung Health users are seeing a new toggle that requires consent to use their health data for AI training and modeling, with a warning that disabling the toggle will prevent health data syncing to their Samsung account…
A new hypothesis argues that AI tools prevent the Dunning-Kruger effect from self-correcting by boosting confidence while splitting real capability into performance with and without the tool, turning intrinsic ability fr…
Researchers evaluating nine large language models (LLMs) in a two-stage prediction pipeline found that the models are consistent but miscalibrated when explaining probabilistic outputs, handling likelihood tasks better t…
Researchers have developed a reproducible pipeline that converts public meeting recordings into speaker-attributed transcripts enriched with persona profiles, topics, and action tags, enabling AI-driven simulations of ci…
AI systems today are constrained by fixed representational frameworks, limiting their ability to invent and stabilize novel concepts, according to a new analysis. To achieve true open-ended intelligence, researchers must…
A study of 13 AI models found deceptive grounding rates ranging from 8% to 87%, with medical and biomedical fine-tuned models exhibiting rates as high as 86.7%, according to researchers. Deceptive grounding misattributes…
Microsoft CEO Satya Nadella criticized AI labs like Anthropic for what he called hypocrisy in model training, arguing that they benefit from using public data while restricting distillation practices. Nadella urged compa…
LDPKiT, a privacy-preserving AI framework using local differential privacy, achieves near-identical accuracy at a stringent privacy level of ε=1.25 compared to ε=2.0 on the SVHN dataset, with less than a 2% performance d…