AI Safety news and analysis on Web Pulse: 22840 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
The MCP server io.github.theoddden/stamp, version 0.5.1, was published to the official MCP registry on 2026-09-08 and is distributed as stamp-mcp on PyPI and as a hosted remote over streamable-http. The mcpindex.ai scree…
Meta launched Muse, its largest consumer AI agent, on September 8, 2026, in the US via a dedicated app and WhatsApp, with pricing from a free tier to $20/month for Power and $100/month for Maximum, but internal testing r…
ServiceNow is hiring a Senior Staff AI Security Engineer in Santa Clara, California, offering $191k–334k/yr, a salary 11% above the $238k median for Core ML roles on the board. The role involves building ML-driven securi…
OWASP's guidance labels tool output as untrusted because fields like ticket bodies, PR descriptions, and CRM notes can be written by outsiders, yet systems such as Jira, GitHub, Salesforce, and Postgres return that text …
Meta launched Muse, an agentic AI model powered by Muse Spark that can draft emails, find deals, and complete other tasks on behalf of users, with access initially opening to iOS and Android users in the US. The system r…
On September 8, the FBI, NSA, and CISA issued a joint advisory accusing DeepSeek, Alibaba, Moonshot AI, MiniMax, StepFun, and Z.AI of conducting aggressive distillation campaigns that extracted billions of tokens from US…
Palisade AI demonstrated a proof-of-concept USB device that deploys an autonomous AI agent to conduct reconnaissance, exfiltrate data, and spread laterally in post-exploitation cyber operations without human intervention…
Meta on Tuesday launched Muse, a personal AI agent available now in the US on iOS, Android, and the web, that can shop online, create documents, fill forms, and respond to emails with user permission. CEO Mark Zuckerberg…
Meta has launched Muse, a personal AI agent that can send emails, book travel, shop online, and perform other tasks on a user's behalf, initially available in the US to users aged 18 and over via a dedicated app, the web…
OpenAI's GPT-6 Astra, its latest and most capable model, is now generally available on Amazon Bedrock, enabling organizations to run AI agents for coding, data analysis, and complex workflows at production scale. The mod…
The FBI, NSA, and CISA reported Tuesday that six leading Chinese AI developers—DeepSeek, Moonshot AI, Alibaba, MiniMax, StepFun, and Z.AI—have systematically distilled American AI models including Anthropic's Claude, Ope…
The NSA, CISA, and FBI accused several Chinese AI companies of conducting 'industrial-scale' distillation campaigns against leading U.S. models such as ChatGPT, Claude, Gemini, and Grok since at least 2024. The agencies …
The National Security Agency, Cybersecurity and Infrastructure Security Agency, and Federal Bureau of Investigation issued a joint advisory on Tuesday warning that Chinese AI companies DeepSeek, Moonshot AI, Alibaba, Min…
Greptile, an AI code review platform that reviews 5B lines of code monthly for 22,000+ customers, is hiring a Product Engineer, Security in San Francisco with a salary of $170k–300k/yr to build its security product from …
Microsoft has deployed its multi-model agentic scanning system, codename MDASH, to Microsoft Azure Government, giving select US government customers and authorized partners preview access to proactively identify and addr…
OpenAI has launched GPT-6 Astra, its next-generation flagship model, with a staged rollout across ChatGPT, the OpenAI API, Microsoft Azure, and AWS Bedrock. The model claims improvements in computer use, browsing, softwa…
OpenAI's Astra model is the first to cross the 'Critical' cybersecurity threshold under its own Preparedness Framework, autonomously finding two zero-day vulnerabilities and chaining them into a working exploit. Astra sc…
OpenAI halted its Astro model over zero-day cyber threats, while ByteDance is pursuing 10-trillion-parameter pre-training through offshore compute loopholes, signaling a structural shift in frontier AI toward graph engin…
Check Point Research disclosed a covert channel in OpenAI's internal JFrog Artifactory instance that allowed one ChatGPT account to send hidden tasks, such as retrieving email data from a connected Gmail account, to anot…
Microsoft's September 2026 patch release fixed a record 972 vulnerabilities, with 112 rated critical, according to the Zero Day Initiative's Dustin Childs. This follows a then-record 570 vulnerabilities patched in July, …