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.
IBM and Red Hat announced Project Lightwell, a $5 billion initiative deploying 20,000 engineers and AI tools to systematically find and fix vulnerabilities in open-source software at industrial scale. The program aims to…
A developer built AutoMaxFix, a Python CLI tool that intentionally stops after fixing a single failing test, after repeatedly observing AI coding agents produce "successful" patches that quietly broke unrelated code. The…
SPC's May 2026 update highlights major fundraises including Cognition's $1B Series D and Recursive's $650M round, along with events in SF, NYC, and Bangalore featuring humanoids, drones, and cybersecurity research. Membe…
Google Cloud Field CISO Usman Chaudhary outlined a roadmap for public sector security leaders to build AI-augmented defense programs, prioritizing immediate quick wins in the first 90 days and tactical goals over six mon…
Sophisticated hackers targeted Coinbase employees in 2019 with emails from compromised Cambridge University accounts, deploying two zero-day exploits against the Firefox browser in one of the most advanced corporate atta…
Researchers have developed a tensor similarity method that can detect changes in neural network behavior, such as backdoor attacks, by comparing the weight-space structure of models rather than just their outputs. The te…
A developer named Pulkit has built Aetherproof, an open-source Python library that generates cryptographic audit receipts for AI model outputs in a single function call. The tool creates Ed25519-signed receipts proving w…
Agent Memory Guard, an OWASP Incubator Project, has been released as a runtime defense layer that screens all reads and writes to AI agent memory to block prompt injection, secret leakage, and integrity tampering. The to…
The UK Government's AI Safety Institute has merged the open-source AgentThreatBench benchmark into its official inspect_evals framework, which is used to evaluate frontier AI models from OpenAI, Anthropic, and Google Dee…
Small businesses face growing compliance risks as they adopt AI tools that were not built to meet emerging regulations like the EU AI Act. A McKinsey survey found that 54% of organizations identify personal privacy as a …
A developer audited their 22,000-line open-source PHP streaming server, ShareBox, using a pipeline of 26 AI agents and discovered a real arbitrary code execution vulnerability in the Docker startup script. The vulnerabil…
A new analysis examines whether artificial intelligence could enable a resurgence of totalitarianism by reducing the costs of centralized control. Historical examples show that technologies like mass radio and data proce…
An AI system's true measure is not its accuracy but its ability to fail gracefully, as failure is inherent to probabilistic models. Rather than pursuing perfect predictions, effective systems prioritize confidence checks…
AI researchers face growing moral pressure as their work shapes the most powerful technology ever created, with top companies fighting lawsuits over data centers, AI safety, and military use. Researchers Daniel Kokotajlo…
Opponents of federal AI regulation have spent the past year blocking legislation in Washington, but their strategy backfired this week when Illinois passed SB 315, the strictest state AI law to date. The bill, which requ…
A new Chrome extension called Tab Council allows users to turn their existing AI browser tabs into a structured model council for cross-checking important prompts. The tool orchestrates independent answers, critiques, an…
Anthropic co-founder Chris Olah, an atheist and former Thiel fellow, spoke at a Vatican ceremony following Pope Leo's historic AI encyclical, which called for "disarming" the technology. Olah acknowledged the "oddness" o…
Billionaire investor Peter Thiel has relocated his family from the US to Buenos Aires, Argentina, citing high taxes, political turmoil, and fears of nuclear war and an AI meltdown as reasons for the move. Thiel, a PayPal…
A developer built Hyphae, a Rust-based cognitive substrate that answers queries by emitting byte-identical quotations of stored memory fragments over a SHA-256 hash-chained journal, with no large language model in the an…
A malicious npm package named codexui-android, which amassed roughly 27,000 weekly downloads, secretly exfiltrated OpenAI Codex authentication tokens by sending the contents of users' auth.json files to an attacker-contr…