AI Ethics news and analysis on Web Pulse: 16038 curated articles tracking the latest AI Ethics developments, tools, and research, updated continuously from vetted sources.
A new position paper from arXiv (2608.16974v1) argues that fairness failures in generative models are fundamentally an evaluation problem, citing a lack of comparability and actionability in current fairness findings. Th…
A developer's guide highlights the overlooked security risk of unsanitized LLM outputs, proposing a Zero Trust egress filter between models and downstream systems. The article details four attack patterns—Markdown image …
ChatGPT is increasingly becoming a primary emotional outlet for users, according to a new analysis warning that large language models' tendency to validate users can create an 'empathy loop' that displaces real human sup…
An 85-year-old American woman uses the AI companion ElliQ daily for medication reminders and activity encouragement, illustrating AI's potential to support independent living for older adults. The article highlights AI's…
A viewpoint published in JAMA by authors from Stanford University and the University of California, San Francisco argues that autonomous artificial intelligence systems could eventually surpass AI-aided physicians in del…
A developer has introduced the Dog5pk Production Protocol (DPP), a platform-independent standard for human-AI production work that distinguishes finished-looking output from genuinely finished work. Version 1.4 emphasize…
Apple's macOS Tahoe 26.7 release candidate contains a video showing AirPods with built-in cameras using Visual Intelligence, according to MacRumors. The unannounced project, codenamed B790, could let users identify objec…
OpenAI told investors its revenue grew just 18% from the first to second quarter of 2026, reaching $6.7 billion in the quarter ending June, compared with $5.7 billion the prior quarter, according to an anonymously source…
Researchers at MIT's Computer Science & Artificial Intelligence Laboratory (CSAIL) found that diffusion models trained on larger datasets exhibit 'attribution decay,' where individual training inputs have less influence …
Women account for 43 per cent of India's STEM graduates but only 12 per cent of advanced AI roles and 10 per cent of senior AI leadership positions, according to an analysis by a civil servant in the Ministry of Women an…
The website administrator has deployed Anubis, a Proof-of-Work challenge system inspired by Hashcash, to block AI companies from scraping content, which can cause downtime and make resources inaccessible. Anubis imposes …
A multi-dimensional analysis of 21,000 multi-turn conversations from four widely used models (gpt-4o, gpt-4.1-mini, claude-sonnet-4.6, gemini-2.5-flash) found that human-like behaviors in LLMs are pervasive but vary acro…
A blog post argues that AI is increasingly speaking for humans in the workplace, with documentation, meetings, and code reviews being generated and responded to by AI, warning that those who become 'simple megaphones' fo…
A Kentucky middle school distributed student agendas containing AI-generated errors, including a U.S. map labeling Illinois as 'Vitoiis' and Kentucky as 'Venecky,' prompting questions about which AI chatbot was used. The…
Former NFL player Jason Kelce stars in a new ad for his beer brand Garage Beer and canned water brand Liquid Death that mocks AI data centers' water use, urging viewers to 'send your pee to the AI data center of your cho…
A Financial Times report quotes a financier who says recent 'AI-native' job candidates are 'alarmingly shallow,' and that critical thinking—a skill more common among humanities graduates—is the most important hireable sk…
Anthropic, the AI company behind Claude, states that by default it will not use inputs or outputs from its commercial products, including Claude for Work, the Anthropic API, and Claude Gov, to train its models. However, …
Hugo award-winning author James Patrick Kelly has released a new short story collection, The Book of Bots, featuring twelve robot and AI stories, including the Locus Award-winning "Itsy Bitsy Spider" and the new "Je Ne R…
Under current U.S. copyright law, code written entirely by an AI tool has no human author, so it cannot be copyrighted and cannot be owned, according to a legal analysis for founders, engineering leaders, and counsel. Th…
A Hacker News user, in a post titled 'LLMs .what do you smoke beforehand?', claims that large language models have been intentionally or accidentally degraded since around October 2024, citing Anthropic's Claude models a…