AI Safety news and analysis on Web Pulse: 12901 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
Anthropic disabled its Fable 5 and Mythos 5 models for all customers after the US government issued an export-control directive requiring suspension of access for foreign nationals. Cohere chief AI officer Joelle Pineau …
Anthropic's Claude AI service experienced an outage on Tuesday, with over 2,000 user error reports on Downdetector. The company implemented a fix by 2 p.m. ET, though errors persisted for Claude Opus 4.8 and Haiku 4.5 mo…
The Department of Justice and Mississippi moved to intervene in an NAACP lawsuit against Elon Musk's xAI over alleged unlawful operation of methane gas turbines near Memphis, arguing that xAI's AI model Grok is vital to …
Microsoft released the Agent Governance Toolkit, a stateless policy engine built on Cedar that intercepts agent actions at sub-millisecond latency and covers all 10 OWASP Agentic Top 10 risks. However, the toolkit does n…
Anthropic launched self-hosted sandboxes in public beta and MCP tunnels in research preview for Claude managed agents, enabling regulated industries to execute tools and access private services within their own infrastru…
Google unveiled Android 17 on May 12, introducing post-quantum cryptography with the ML-DSA algorithm to secure devices against future quantum threats, while major AI features like the Rambler assistant and AI widgets ar…
Autonomous coding agents with shell access, tokens, and no human oversight break traditional DevSecOps by acting at machine speed on external inputs, bypassing code review and CI gates. A developer argues that the contro…
Anthropic CEO Dario Amodei responded to a video about an Iranian girl school strike, addressing the situation in a public statement.
Recent advances in quantum computing have shortened the timeline for breaking RSA encryption to within five years, prompting Google to call for accelerated adoption of Post-Quantum Cryptography (PQC) by 2029. The risk st…
Anthropic took its Claude Fable 5 and Mythos 5 AI models offline after a US government export-control directive barred foreign nationals from using them, citing national security risks. Experts warn that similar dangerou…
AI tools are enabling a surge in cyberattacks, with Palo Alto Networks reporting a fourfold increase in daily attacks from 2024 to 2025. Hackers are using AI-enhanced viruses that adapt to avoid detection and automating …
A couple planning to donate at least half their assets over their lifetime favors GiveWell and animal welfare over AI safety, despite having a comparative advantage in AI safety giving that could be 10 times more cost-ef…
A developer built Pramagent, trust middleware for LLM agents that enforces deterministic guarantees outside the model. The system wraps any LLM provider with ten independent layers for compliance, safety, reliability, an…
UK Prime Minister Keir Starmer's request for a carveout from US export controls on Anthropic's advanced AI models Fable 5 and Mythos 5 was denied by the Trump administration, which deemed exemptions for G7 allies 'comple…
The US government imposed export controls on Anthropic's advanced AI models Fable 5 and Mythos 5 on June 12, 2026, restricting access to US nationals due to national security concerns, sparking European backlash and call…
Ken Walger announced the release of two new components for the Sovereign SDK: sovereign-sieve and sovereign-ledger. These tools extend the Sovereign Systems Specification from a network gateway to local silicon, enabling…
Google was named a Leader in the 2026 IDC MarketScape for Worldwide SIEM Vendor Assessment, recognizing its Google Security Operations platform. The platform integrates Mandiant expertise, AI agents, and Gemini models to…
HPE announced enhancements to its AI Factory portfolio at HPE Discover 2026, introducing the HPE ProLiant Compute DL394 Gen12 powered by NVIDIA's Vera CPU and adding support for the NVIDIA Agent Toolkit to operationalize…
Insinuator.net disclosed a vulnerability in Firefox's AI sidebar that allows prompt injection to exfiltrate user email addresses. The browser sends page titles and selected content to third-party chatbots like Claude and…
Zscaler published a blog post on June 16, 2026, advocating Zero Trust as the security model for AI agents in enterprises. The post argues that agentic systems expand the trust surface and require controls like verified i…