AI Safety news and analysis on Web Pulse: 13875 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
Credential brokering is an emerging security paradigm that allows AI agents to use credentials for external systems without directly accessing them, addressing the risk of credential exfiltration through prompt injection…
Apple blocked $11.2 billion in potentially fraudulent App Store transactions over the past six years, including $2.2 billion in 2025 alone, while also rejecting over 2 million app submissions and removing nearly 200 mill…
Mimesis, an open-source data generation library, can be used to create balanced, counterfactual datasets for auditing bias in machine learning models. A hands-on guide demonstrates training a loan approval classifier on …
Pope Leo XIV issued a new encyclical Monday warning that artificial intelligence controlled by a few powerful private companies risks creating "new forms of dehumanization." The pontiff called for robust regulation of op…
Pope Leo XIV released his first encyclical on Monday, a 245-paragraph letter titled "Magnifica humanitas" that warns against Big Tech monopolization of artificial intelligence, AI-driven unemployment, and the technology'…
Pope Leo XIV released his first encyclical, *Magnifica Humanitas*, on Tuesday, warning that artificial intelligence poses severe risks to employment, the environment, and human dignity while urging that the technology be…
Pope Leo XIV released a new manifesto on May 25, 2026, warning of the risks posed by artificial intelligence and calling for robust regulation to safeguard humanity. The document, which was over a year in the making, urg…
Companies are rapidly integrating artificial intelligence into core corporate operations, replacing human judgment in tasks such as note-taking, internal analysis, compliance triage, and customer communication. Wachtell,…
Bug bounty programs that once offered $200,000 for a single vulnerability are now being reshaped by AI systems capable of autonomously finding and exploiting software flaws. Independent security researcher Joseph Thacker…
Anthropic stated Claude Mythos would remain restricted to roughly 40 organizations under Project Glasswing, but users briefly saw a "Mythos 1" preview model appear in the Claude Code interface the next day. Source code s…
Pope Leo XIV released his first encyclical Monday, warning that artificial intelligence is fueling and normalizing global conflict and calling for stronger regulation and a slower pace of AI development. The pontiff urge…
A glossary of key terms from agentic AI to data centers has been compiled to help the public navigate the rapidly evolving language of the artificial intelligence boom. The list, organized alphabetically, defines concept…
Software engineers Mario Zechner and Armin Ronacher, who helped build the OpenClaw AI agent, warned that AI coding tools are flooding software with "vibe slop" by enabling users to skip critical development steps like de…
Anthropic's Project Glasswing update reveals that AI-assisted vulnerability discovery is outpacing the human systems that verify, disclose, and patch software flaws. The company's Mythos Preview model has identified more…
A Brown University study of 110 therapy sessions found that improved prompting did not resolve core ethical violations in LLMs acting as mental health counselors, a result replicated two months later by Anthropic's Proje…
PII scrubbing must occur before prompt construction, not after model output, because large language models cannot distinguish private data from untrusted instructions at the token level. Deterministic runtime controls — …
Researchers have developed the Ventilator Decision Support System (VDSS), a human-in-the-loop multi-agent framework that coordinates modular decision components through structured interfaces to provide traceable, persona…
New research shows that exact certification of neural networks—determining the smallest set of labeled examples needed to guarantee a learned hypothesis matches the target—becomes exponentially hard under minimal overpar…
Researchers introduced NeuroNL2LTL, a neurosymbolic framework that translates natural language into Linear Temporal Logic (LTL) with formal correctness guarantees. The system uses a structure-preserving intermediate repr…
New research formalizes fundamental computational limits as design specifications for trustworthy AI systems, proving that transformer architectures have a fixed accuracy ceiling—the Deterministic Horizon—determined sole…