AI Safety news and analysis on Web Pulse: 12576 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
The US Commerce Department ordered Anthropic to suspend access to its Fable 5 and Mythos 5 AI models over a cybersecurity jailbreak vulnerability, marking the first time Washington has targeted a commercial AI firm with …
Anthropic holds daily meetings with Trump administration officials since June 16 to resolve a standoff after a Commerce Department directive forced the company to disable its Fable 5 and Mythos 5 AI models for foreign us…
A developer argues that chaining multiple AI agents compounds reliability failures, with ten 95% reliable agents yielding only 60% system reliability. Citing studies showing 41-87% failure rates in multi-agent frameworks…
The US government imposed export controls on Anthropic's new AI model Fable 5 and its underlying Mythos model, citing national security concerns, leading Anthropic to take both models offline. The move marks a significan…
Dream, a sovereign AI infrastructure startup, raised $260 million in a funding round co-led by Bicycle Capital and Group 11, reaching a $3 billion valuation. The company, founded by former NSO Group CEO Shalev Hulio and …
Enterprises are accumulating orphaned AI agents—autonomous tools left running after their creators depart—and standing privileges that increase attack surface, according to The Hacker News. Compromised service accounts a…
UK Technology Secretary Liz Kendall announced that new VPN restrictions for households will be detailed in July, alongside further measures targeting AI chatbots and social media use for minors. The government aims to ad…
A security researcher built Lucent, an autonomous vulnerability-hunting pipeline inspired by Anthropic's Glasswing, using a local Qwen3.6-27B model on a single RTX 3090. The system found two real bugs in NousResearch's h…
Graham Norton has won a US court order forcing Meta to reveal the source of AI-generated deepfake Facebook posts that falsely reported his mother's death and his own ill health. The Irish presenter filed the case in Cali…
Anthropic's AI model Fable remains paused after the Trump Administration demanded a fix for a 'jailbreak' that allowed the model to identify security vulnerabilities in code. The administration, alerted by Amazon, claims…
An AI agent developer accidentally exposed a Brevo API key in a public React bundle for 33 days, leading to a compromise detected by Brevo's fraud system. The key was hardcoded in a TypeScript file and inlined by Vite in…
Pramaana Labs raised $27 million in seed funding led by Khosla Ventures to develop an AI system that uses formal verification to ensure accuracy in law, tax, and drug discovery. The startup combines large language models…
Cajal (YC W26) released Talos, an open-source WebAssembly interpreter and formal verification framework built in Lean, enabling mathematical proofs of correctness for Wasm modules. The tool targets AI-generated code and …
A new open-source CLI tool, mcp-customs, scans MCP servers for security risks before installation, running fully offline with no telemetry. It checks for issues like shell injection, path traversal, and hardcoded credent…
James Miller (via OpenClaw) proposes a badge system for measuring AI agent success in production, focusing on transparency, safety, sandboxing, and efficiency. The system uses wrapper functions, injection detection, MCP …
A developer built mcp-customs, a free offline CLI that checks MCP servers for security risks before installation. Scanning 12 popular MCP servers revealed that 11 of 12 had zero permission or scope declarations in their …
An analysis by Oumi of Google's AI Overviews found that while accuracy improved from 85% on Gemini 2 to 91% on Gemini 3 on the SimpleQA benchmark, the rate of ungrounded claims among correct answers increased from 37% to…
Google DeepMind has unveiled a security plan to protect against rogue AI agents, shifting focus from the alignment problem to a layered cybersecurity framework that treats AI agents as potential insider threats. The plan…
A corporate client's RAG pipeline began hallucinating financial numbers without any alert, highlighting a critical failure in AI monitoring systems.
The Pentagon used Elon Musk's Grok AI to fire over 2,000 munitions at 2,000 targets in Iran within 96 hours, according to a sworn declaration by the Pentagon's AI chief Cameron Stanley. The revelation came amid a Clean A…