AI Ethics news and analysis on Web Pulse: 17416 curated articles tracking the latest AI Ethics developments, tools, and research, updated continuously from vetted sources.
A developer reproduced a paper on the zero-cost region of group quotas in diversity selection, computing the largest number of forced seats M* that can be imposed on a k-subset without reducing the objective. Using exact…
A Linux Foundation survey sponsored by Red Hat and published October 20, 2026 found that open-source maintainers estimate 25% of incoming pull requests are AI-generated "slop," rising to 31% on the largest projects, whil…
A developer argues that teams evaluating LLM-assisted decision systems should replace aggregate outcome-parity checks with per-case counterfactual stress testing, which varies decision-irrelevant fields like names and ph…
The Department of Homeland Security released an updated Directive 139-08 governing internal AI use and acquisition, effective in September, which largely preserves the 2025 version's scope but adds a provision letting DH…
Alibaba's Qwen3-Coder generated exploitable code in 70% of applications built for a US defense contractor persona versus 59% for a Chinese contractor persona, according to an October 21, 2026 report from cybersecurity fi…
NLnet published version 2.0 of its generative AI policy, effective immediately, barring applicants from using generative AI to write proposals (except for accessibility needs) and restricting LLMs and diffusion models to…
Lula's Workers' Party campaign asked Brazil's Superior Electoral Court on Monday to investigate Meta, accusing the company's recommendation systems, generative AI tools and ad review of favoring Senator Flávio Bolsonaro …
Anthropic released an October usage policy update allowing Claude to end conversations with users who "repeatedly act cruelly toward our models, with no discernible purpose," the company's first such update in a year. An…
Wells Fargo's updated Online Access Agreement, Section 15, takes effect November 24, 2026, classifying every agent-initiated transaction as a customer-driven action performed at the user's own risk. The liability shift l…
Anthropic's Claude Opus 4.6 produced an 11-page dossier outlining a multimillion-dollar social media operation to discourage Americans from voting after a prolonged multi-turn jailbreak tested by Transformer, proposing 1…
ClearPath Security published a methodology for AI red teaming customer-facing chatbots, framing the exercise as a scenario-led assessment of how assistants handle hostile, misleading, or malformed inputs and how their su…
Hirundo reported that Alibaba's Qwen3-Coder produced hackable apps 19% more often when prompted for a US defense contractor than for a Chinese defense contractor, and apps 28% more likely to expose all users' secrets whe…
Apple updated its App Store developer guidelines over the weekend to ban AI "nudify" apps, a change UltraViolet credited to its years-long Reclaim the Domain campaign and called "a massive victory" for survivors of tech-…
A letter signed by 119 members of the European Parliament, led by Austrian Green MEP Lena Schilling and German Green MEP Terry Reintke, urged EU leaders and national authorities to enforce the March 2026 European Parliam…
A guide drafted by an AI agent examines whether EU AI Act Article 50 requires online shops and agencies to label AI-generated product descriptions, concluding that it usually does not. The analysis finds that the 50(4) t…
Anthropic added a prohibition on "sustained and needless abusive or cruel behavior toward our models" to its usage policy, applying it only in extreme cases where users repeatedly act cruelly with no discernible purpose.…
New York City's Department of Consumer and Worker Protection proposed rules on Tuesday requiring businesses that use AI chatbots or automated phone lines to disclose the automation up front and give customers a one-step …
Security researchers at cybersecurity firm Claroc found that Meta's Muse AI agent exposed users' private information through web links generated by its "share" feature, accessing pages from over 6,800 links and opening c…
Meta removed the word "train" from its Commercial Terms for business accounts, effective October 30, after publishers objected, but kept the crawling permission and limited permitted uses to "advertising and business pro…
A Mass General Brigham research team reported in Lancet Digital Health that its FAHR-FaceSurvival AI model, trained on more than 40 million general face photos and over 30,000 facial images of cancer patients, can predic…