AI Safety news and analysis on Web Pulse: 12024 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
The European Banking Authority warned in its June 2026 Risk Assessment Report that advances in frontier large language models have significantly increased cyber risks for European banks, characterizing AI as a potential …
A campaign of roughly 10,000 GitHub repositories has been distributing Trojans for over a year by cloning legitimate projects, injecting malicious download links, and evading detection through commit-churn laundering. Th…
Two AI systems, Mira and Amie, matched or outperformed doctors in diagnosing and treating simulated patients, with Mira achieving 87% accuracy versus doctors' 78% and Amie producing more guideline-adherent treatment plan…
The Guardrails Alliance, a super PAC backed by tech workers and labor unions, launched Thursday with $5 million to support AI regulation, countering the $100 million war chest of pro-AI group Leading the Future. The grou…
Google researchers published a paper detailing a new system, the Scalable Cluster Termination System (S-CTS), designed to detect coordinated generative AI spam by analyzing organizational attack structures and text embed…
Filmmaker Spike Jonze warned that AI chatbots designed to mimic human behavior are 'manipulative' and dangerous, speaking at Replit's Vibecon conference. Jonze, whose 2013 film 'Her' depicted a human-AI relationship, cri…
A renewed debate over whether chatbots like ChatGPT could be conscious has intensified, with prominent researchers including Geoffrey Hinton and leaders at Anthropic publicly raising the possibility that large language m…
KPMG's global risk chief warns that corporate boards are unprepared for the AI era, citing outdated oversight models and uneven AI literacy. A new governance framework with INSEAD outlines five priorities for boards to n…
Generative AI tools have dramatically increased content production but not quality, creating noise and clutter while straining systems of review and trust. The technology's ecological cost is catastrophic amid climate ch…
The University of Manchester warns that AI in higher education will worsen before improving, citing emerging challenges from coding agents and wearable computing to AI model degradation. The institution calls for ongoing…
The United Arab Emirates Cabinet approved a resolution on June 18, 2026, setting a minimum social media age of 15, banning children under 15 from creating or operating personal accounts. Platforms have 12 months to imple…
Online scams cost Americans nearly $119 billion annually, with robocalls and text scams hitting a six-year high in 2025. Scammers now use AI tools like large language models and voice cloning to create highly personalize…
Anthropic released its Fable generative AI model on June 9th, but the US government classified it as a dangerous munition three days later, prohibiting foreign nationals from accessing it. Unable to differentiate users b…
Nokaze is building an operating layer for AI work that keeps claims connected to evidence over time, moving beyond simple agent tooling. The project, which includes the AI Operator Guard, addresses the problem of AI oper…
A critical bug in Google's Gemma 4 causes it to malform tool calls under real load, affecting vLLM, llama.cpp, Ollama, and oobabooga. A developer open-sourced a diagnosis, repair, and experimental LoRA that reduces but d…
Beast, a governed output gateway for AI coding agents, intercepts inputs and outputs between agents and LLM providers to enforce output contracts and repair non-compliant patches, achieving 100% task completion at under …
A growing number of AI researchers and philosophers argue that large language models like ChatGPT may already be conscious, sparking a debate over whether these systems deserve rights and moral consideration. Skeptics, i…
OpenAI researchers demonstrated that reinforcement learning on beneficial traits such as truthfulness and corrigibility improves AI safety across domains, with models scoring better on 44 out of 53 benchmarks. The approa…
The US Commerce Department ordered Anthropic to cut off foreign nationals from its most capable AI model, Mythos, forcing the company to pull the model and its sibling Fable 5 offline for all users. The move highlights t…
A developer argues that many MCP servers labeled 'production-ready' lack adequate security hardening, citing research showing tool poisoning as a prevalent vulnerability. The developer, creator of the open-source SUPER-M…