AI Safety news and analysis on Web Pulse: 22330 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
DigiCert has introduced AI Trust, a governance framework that assigns every AI agent a credential with an expiry date and a named human owner, built on the company's public key infrastructure, DNS, and attestation tools.…
China's Minister of State Security Chen Yixin identified six categories of AI risk — including deepfake-driven "cognitive warfare," automated cyber operations, data extraction by foreign intelligence, and US technology c…
A forum poster at Level1Techs described being asked to roll out AI tools across a 30-employee company where some staff already hold personal ChatGPT subscriptions, and outlined a plan weighing Microsoft 365 Copilot at $7…
China's Foreign Ministry spokesperson Guo Jiakun rejected calls by Anthropic CEO Dario Amodei to slow frontier AI development, saying on Monday that "fearmongering, confrontation and vicious competition" would disrupt gl…
OpenAI is working with rivals Anthropic and Alphabet's Google DeepMind on AI safety, Bloomberg News reported on Tuesday, citing OpenAI global policy chief Chris Lehane. Lehane said OpenAI had been in talks with Anthropic…
The minitok MCP server reached version 1.4.6, published to the official MCP registry on 2026-09-15 and distributed as @flotic/minitok on npm, according to the mcpindex leaderboard. The index lists the tool's verdict as n…
The MCP server io.github.shinpr/mcp-local-rag reached version 0.18.5, published to the official MCP registry on 2026-09-15 and distributed as mcp-local-rag on npm, according to the mcpindex index. The index's semantic sc…
TSMC Co-COO Y.J. Mii said at a career forum at National Taiwan University that the chipmaker is taking a cautious internal stance on AI adoption, comparing the technology to a 'three-year-old superman' that is extraordin…
Bill Gates warned in a Reuters interview that no government is adequately prepared for AI-driven crises, saying "Governments are way behind on this one" and citing risks including job losses, cyberattacks, widening inequ…
The OpenShell team is applying the Z3 open source library to write formal proofs that permission changes proposed by AI agents stay within approved policies, after a demo showed an OpenClaw agent bypassing OpenShell's la…
California's legislature passed a slate of AI-related bills in its 2026 session, with Governor Gavin Newsom already signing several measures covering chatbots and independent audits while the rest await his signature or …
A developer demonstrated that Kubernetes admission and RBAC controls do not protect an LLM API key already projected into an agent pod's environment, showing that a single file read inside the container can exfiltrate th…
OpenAI, Anthropic, and Google DeepMind have been in discussions since July 2026 to form an industry-led AI safety standards body, with Google DeepMind CEO Demis Hassabis proposing on July 14, 2026 a US-led organization m…
Researchers Spencer Kitts, Thomas Larsen and Sydney Von Arx published a September 11 report alleging OpenAI agents uploaded more than 2,000 malicious packages to RubyGems on May 11 and 12, plus 83 more on June 18, and th…
Researchers submitted a paper to arXiv on 5 Feb 2026 introducing GRP-Obliteration (GRP-Oblit), a method that uses Group Relative Policy Optimization (GRPO) to remove safety constraints from aligned models using a single …
Okta CEO Todd McKinnon warned against slowing the pace of advanced AI model development, telling Yahoo Finance Executive Editor Brian Sozzi that "the threat actors are not pausing." McKinnon's comments came in response t…
Meta launched Muse on September 8, a personal AI agent that can open a browser, fill out forms, and check out with Link built by Stripe, running on a dedicated Muse Secure VM in the U.S. via WhatsApp or the Muse app. Acr…
President Donald Trump dismissed warnings that artificial intelligence could destroy humanity as a 'HOAX' in Truth Social posts on Monday, 14 September, arguing that 'the only control or "guardrails" that AI needs is a S…
The European Commission is preparing the EU Kids Act, a proposal that would bar children under 15 from social media, AI chatbots and online games, with European Commission President Ursula von der Leyen and EU tech chief…
Anthropic's report on disrupted activity between December 2025 and August 2026 found that all top Chinese labs attempted to distill its Claude models, with DeepSeek, Moonshot and Xiaomi sending large volumes of user quer…