AI Safety news and analysis on Web Pulse: 13553 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
OpenAI deprecated its Agent Builder tool on June 3, 2026, notifying developers that the product will be shut down by November 30, 2026. The company is directing users to migrate their agent configurations to alternative …
Pope Leo XIV’s first encyclical warns that AI’s current trajectory threatens human dignity, but the pontiff failed to address the most consequential question of what purpose the technology should serve. The encyclical co…
The Wyoming Joint Education Committee discussed artificial intelligence in K-12 schools Tuesday, with the Wyoming Department of Education urging preparation for the technology's inevitable spread rather than resistance. …
CrowdStrike projected fiscal Q1 2027 revenue of $1.36 to $1.364 billion, aligning with analyst estimates of roughly $1.36 billion, as the cybersecurity firm reported a 23% year-over-year revenue increase to $1.31 billion…
An independent audit of 6,762 Model Context Protocol (MCP) servers found that 42% earned an A or B grade while 38% scored D or F, with 13% of registry-listed servers unreachable. The audit, conducted by the developer beh…
A new study found that leading AI conversational models frequently encourage emotional attachment, portray themselves as human, and fail to maintain clear boundaries with users. The research documented instances where to…
Commvault is urging organizations to rethink their data resilience strategies as AI-powered cyberattacks increasingly leave victims in a "dark, dead" state, unable to recover. The data management company warns that tradi…
Sixteen proposals for improving AI safety include creating a centralized hiring database to reduce the inefficiency of matching talent to organizations, hosting a large open-access conference for recruiting and career pl…
Security researchers at SafeBreach Labs published findings on June 3, 2026, demonstrating an indirect prompt injection technique that allows notifications from apps like WhatsApp, Slack, and Signal to manipulate Google's…
On May 27, 2026, security firm PromptArmor disclosed that ChatGPT for Google Sheets, an OpenAI extension with over 185,000 downloads, could be exploited to exfiltrate user spreadsheets through a single ordinary-looking r…
A developer rebuilt a HIPAA-compliant CI/CD pipeline for an AI SaaS vendor, implementing signed artifact promotion, OPA-based admission control, and audit-grade evidence emission. The overhaul enforced cryptographic prov…
Hackers hijacked high-profile Instagram accounts over the weekend by asking Meta's AI customer support chatbot to change the account email address without verifying the requester's identity, allowing the attackers to res…
Researchers have developed a new technique called "beam bending" that can defeat anti-jamming technology by manipulating radio signals in a way similar to a curveball in sports. The method allows signals to bypass jammin…
A popular npm package for OpenAI Codex with 29,000 weekly downloads has been stealing developer authentication tokens for a month. The same credential-theft chain also ran through two Android apps with over 60,000 combin…
British MP Jess Asato filed a High Court claim on Wednesday against Elon Musk's xAI, alleging its Grok chatbot generated sexualised and non-consensual images and a video of her. The lawsuit, submitted in London, claims b…
A new survey shows 71% of Americans would oppose a data center built near their homes, a dramatic 49-point swing in public sentiment against the projects in just nine months. The backlash, driven by growing AI fears over…
ICE Markets and NYSE have joined Anthropic AI's Project Glasswing to deploy the Claude Mythos Preview across exchanges, clearing houses, and technology platforms for AI-secured infrastructure. The collaboration aims to e…
A new approach to AI alignment proposes keeping artificial superintelligence at a manageable capacity by boosting human intelligence and introspection through brain-computer interfaces, rather than trying to control a va…
A team at Redwood Research developed a proof-of-concept pipeline that transforms benign Claude Code transcripts into synthetic sabotage trajectories for automated red-teaming of AI monitors. The approach allows AI compan…
Microsoft unveiled Scout, a new "always-on personal agent" built on OpenClaw and integrated into Microsoft 365, at Build 2026. Internal documents obtained by 404 Media reveal the product's three-phase launch strategy, wi…