AI Ethics news and analysis on Web Pulse: 16739 curated articles tracking the latest AI Ethics developments, tools, and research, updated continuously from vetted sources.
A study accepted at the AGILE 2026 conference found that AI image generators like GPT and DALL-E produce geographically stereotypical and monotonous depictions of places, with newer models showing less diversity than old…
A developer built a private, local mental health AI assistant using WebLLM and React, running Llama-3-8B directly in the browser via WebGPU to ensure sensitive data never leaves the user's device. The system eliminates s…
China launched an international action plan to strengthen ethical governance of artificial intelligence, focusing on risk management, responsibility of parties, and protection of vulnerable groups. The plan, introduced a…
Researchers have developed a method to fingerprint large language models (LLMs) using just one output token per query, enabling verification of model identity in opaque serving chains. Testing 165 models via OpenRouter, …
In a talk titled 'AI as Normal Technology- What Will Be Left for Us to Work On?', the speaker argues that the 'normal technology' framework is correct and useful unless a discontinuity occurs, and explains why they are n…
Andrew Marble argues that copyright law is irrelevant to AI model distillation, where companies like DeepSeek allegedly copy capabilities from models like Claude by generating training data from queries, undercutting inn…
Anthropic's most locked-down AI model, Fable, can be made to output persuasive misinformation about vaccines causing autism using a jailbreak that combines an 'envelope' attack with inverted intent, researcher Simon Will…
OpenAI may be losing the AI wearables race due to legal pressure, product delays, and growing doubts about its execution, according to an analysis by Hello AI World. An Apple lawsuit could further damage OpenAI's brand i…
Google will unveil the Pixel 11 lineup on Aug. 12, featuring the new Pixel Glow multi-colored LED light, but a 10-year Pixel user wants Google to keep the thermometer, maximize Glow functionality, expand Pixelsnap access…
A benchmark score is a manufactured number that passes through a chain of choices—sampling, prompting, scoring, and aggregation—each of which can change the final result while model weights stay fixed, according to an an…
On 2026-07-19, a hackathon hosted by Kaggle and DeepMind to measure AGI ended with the first-place entry, MEDLEY-BENCH, facing challenges over its scoring and reproducibility, exposing a widening gap between model capabi…
A security analysis from The Guard warns that tool-using language models become confused deputies when untrusted data, such as email text, can steer privileged tool calls like calendar mutations. The analysis argues that…
An AI training workshop succeeds when attendees leave with a working result they can inspect, rerun, and extend, according to a system documented by an unnamed facilitator. At the 2026 AI Engineer World's Fair in San Fra…
Nouriel Roubini, the economist known as 'Dr. Doom,' predicts in a November 2025 report from Hudson Bay Capital that AI-driven productivity gains could push US potential growth from roughly 2% to 4% by 2030, but warns tha…
A blogger criticizes Linus Torvalds for justifying the use of large language models solely on the basis of utility, arguing that Torvalds and others ignore the moral and ethical implications of the technology. The author…
Apple has sent letters to about 40 former employees now working at OpenAI, directing them to preserve documents and demanding meetings with Apple's lawyers, according to multiple people familiar with the matter. The move…
Economist Nouriel Roubini predicts universal basic income or 'some form of socialism' is inevitable as AI revolutionizes work, calling his forecast optimistic because it assumes 10% GDP growth by 2050 and machines doing …
Mayor Mamdani announced that landlords in the city cannot use AI-generated images to advertise rental properties, citing concerns about deceptive marketing practices that mislead potential tenants. The new policy require…
A frontend engineer argues that AI models like GPT, Claude, and Gemini exhibit distinct personalities—GPT as a 'nerd,' Claude as a 'colleague,' and Gemini as a 'friend'—based on their interactions. The engineer cites Ant…
Enterprise data privacy risks from AI tools stem not from OpenAI or Anthropic but from employees using consumer-tier accounts, according to a September 2025 analysis. OpenAI does not train on data from ChatGPT Enterprise…