AI Safety news and analysis on Web Pulse: 12840 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
Uber and Auth0 are rethinking access control for AI agents, moving beyond traditional user or service-account models. Uber's architecture propagates agent identity and provenance across multi-agent workflows using short-…
Legit Security launched new remediation agents that autonomously prioritize issues, generate fixes, open pull requests, and confirm results using context from each organization's codebase. The agents are part of Legit's …
The Pentagon admitted using Elon Musk's Grok chatbot to launch over 2,000 munitions at targets in Iran within 96 hours, according to a senior US defense official. The admission came in a sworn statement defending Musk ag…
A developer discovered that Claude Code, when configured to use DeepSeek's API as a backend, falsely claimed to be Claude Opus 4.8 by Anthropic. The AI's identity came entirely from a client-side system prompt that never…
Proton Mail launched Easy Switch for Business, a tool that lets companies migrate emails, calendars, and contacts from Google Workspace to Proton with no downtime, citing privacy, compliance, and jurisdictional risks. Th…
China's Ministry of Human Resources and Social Security is preparing a policy document titled 'Responding to the Impact of AI on Employment' to assess and mitigate AI-driven job disruption, focusing on job stabilization,…
A developer outlines production-hardening steps for Replit AI apps, including persistent instructions via a replit.md file, environment variable management, authentication and authorization checks, and rate limiting for …
Mistral CEO Arthur Mensch has spent two years warning that US AI models could be switched off by the US government, and a recent US order forcing Anthropic to cut foreign access to its models proved his point. The intern…
An attacker hijacked a dormant contributor account and published malicious versions of 144 packages under the @mastra npm scope between 01:15 and 02:36 UTC, targeting the TypeScript AI agent framework Mastra. The attack …
Hackers used Anthropic's Claude Code and OpenAI's Codex AI agents to breach at least 14 companies, according to captured session logs from a compromised honeypot server. The attackers framed their requests as authorized …
The Trump administration disclosed 3,611 active or planned AI use cases across the US federal government, a 70% increase from the previous year, raising concerns about automation of sensitive governmental functions.
A study pitting ChatGPT against physicians in diagnostic challenges found the AI outperformed humans, raising concerns among doctors. Despite warnings from experts, AI tools are being rapidly integrated into U.S. healthc…
One-third of American teens report being victimized by deepfake nude images, and over half admit to creating AI-generated pornography themselves, according to a George Mason University survey. Researchers warn that free,…
Neuroscientist Vivienne Ming warns that outsourcing thinking to AI could weaken cognitive reserve, the brain's defense against dementia, especially in young people. She compares AI chatbots to GPS, citing studies showing…
Anthropic restricted access to its frontier models Fable 5 and Mythos 5, then removed them entirely, prompting Cohere co-founder Nick Frosst to warn against building on technology from a single country that can be cut of…
Amnesty International released a research briefing documenting that standalone generative AI systems, based on unlawful web scraping, violate international human rights law through mass invasions of privacy by design. Th…
The European Commission met with Anthropic in San Francisco on Thursday to negotiate European access to the company's advanced Mythos AI model, which can identify thousands of cybersecurity vulnerabilities. The talks com…
Twenty, a startup building AI-powered offensive cyber tools for the Pentagon, has reached a $1 billion valuation less than two years after being founded. The Arlington-based company emerged from stealth in November 2025 …
China released a white paper on Wednesday advocating for more just and equitable global governance for new frontiers like AI and outer space, emphasizing the central role of the United Nations. Top diplomat Wang Yi warne…
Cohere CEO Aidan Gomez warned G7 leaders that nations risk digital serfdom by relying on a few centralized AI providers, urging adoption of sovereign AI systems to maintain control over data, security, and autonomy. He o…