AI Ethics news and analysis on Web Pulse: 16085 curated articles tracking the latest AI Ethics developments, tools, and research, updated continuously from vetted sources.
Marketers are increasingly integrating AI tools into their workflows without universal guardrails, according to a Digiday Podcast discussion featuring Evin Shutt, global CEO of 72andSunny, and Isaiah Clay, a recent gradu…
A new book by historian and journalist Chris Patten warns that the rise of digital technology and artificial intelligence threatens liberal democracy, replacing democratic deliberation with calculation and data-driven co…
A 128-GPU cluster built entirely from retired data center GPUs, called DumpsterCluster, can serve LLaMA-70B inference at competitive throughput for $22K versus $600K for an 8-GPU Nvidia B200 system, according to a new ar…
A new survey from arXiv (arXiv:2608.14730v1) proposes a two-dimensional taxonomy for intellectual property protection in visual generative AI, classifying methods by control logic (Information Exposure Control, Generativ…
Researchers introduced EARL (Explanations using Alternative Realities for Reinforcement Learning), a Python library that generates counterfactual explanations for reinforcement learning decisions in self-adaptive systems…
A new arXiv paper (2608.14642v1) comparing four methods for training ethical reinforcement learning agents on the Craftax benchmark finds that while all methods perform similarly on average, they differ sharply per episo…
A new comparative review of AI regulations in the EU, US, and China finds that cross-jurisdictional divergence creates compliance uncertainty for high-risk AI operators, identifying three recurring gaps: weak interoperab…
A new position paper on arXiv (2608.14568v1) argues that AI governance must be built on ISO-like interoperability protocols rather than laws alone, citing the fragmentation caused by the EU AI Act, China's algorithm gove…
A new arXiv preprint (2608.14566v1) argues that evaluations of large language models' moral reasoning focus almost exclusively on the moral value problem—whether outputs align with human values—while largely ignoring the…
A position paper on arXiv (2608.14598v1) argues that medical AI is neglecting real treatment outcomes, relying instead on human opinions and text syntheses, which limits its potential and causes deficiencies in frontier …
A study from arXiv (2608.14630v1) found that large language models (LLMs) induce an average 2.81% rate of harmful decision flips in clinical decision-making, where clinician participants change from correct to incorrect …
Researchers from the Allen Institute for AI found that Dolma, the open training corpus for the OLMo series of large language models, likely contains hundreds of thousands of documents with extremist content and hate spee…
A new study from arXiv (2608.14843v1) finds that academic advisors' writing styles are 39.9% closer in cosine distance to their students' styles than random same-field authors, based on 5,803 authors and 2,501 advisor-st…
A new study from arXiv (2608.14651v1) evaluating open-weight Multi-Modal Large Language Models (MM-LLMs) for disaster assistance found that no model achieves reliable consistency across text and audio modalities, with pe…
A new arXiv preprint (arXiv:2608.15382v1) proposes a causal knowledge graph-based evaluation framework for healthcare large language models (LLMs), tested in a cardiovascular pilot. The framework's integrated grounding c…
TikTok is using an AI clone of a brand founder for advertisements, treating the founder's identity as a set of model weights to generate personalized ads at scale. The platform leverages a few minutes of source footage t…
A candy brand discovered that TikTok created an AI clone of its founder to run ads, according to a video report. The clone was used in advertisements without the founder's consent, raising concerns about unauthorized AI-…
Anthropic announced on August 12 that it would embed an imperceptible watermark in text generated by Claude, prompting some users to object because it would catch them using the AI at work and in classes, according to Te…
A developer voices that AI is draining the joy and meaning from side projects, as new software releases feel like wrappers and LLMs can explain, implement, and optimize concepts in seconds. The author questions whether s…
Epoch AI estimates the effective stock of quality-adjusted public text at roughly 300 trillion tokens, which frontier developers could exhaust between 2026 and 2032, signaling the end of the scrape-first era. Legal and r…