AI Safety news and analysis on Web Pulse: 9281 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
The US government has allowed Anthropic to release its Mythos 5 AI model to approximately 100 trusted US companies and federal agencies, reversing previous restrictions over national security concerns. Meanwhile, a Chine…
Developer Maneshwar is building git-lrc, a free and source-available Micro AI code reviewer that runs on every commit. The project highlights the critical need for data privacy in AI systems, warning that models can memo…
Eric Gilliam proposes that AI philanthropists adopt Warren Weaver's risk-tolerant, concentrated investment strategy to maximize impact, drawing on a blog prize by Dwarkesh Patel and a post by Stripe Climate GM Nan Ransoh…
FAME Knowledge Agent Gateway, an open-source external memory and tool-safety gateway for local coding agents, has been released. It combines a graph knowledge net, project memory, tool-call governance, and a visual workb…
A growing sentiment among tech professionals holds that AI is making them worse engineers, as layoffs and unreasonable demands from leadership compound the chaos of adopting new AI tools. The author argues that no one tr…
Palantir Technologies and Nvidia announced a Sovereign AI Operating System Reference Architecture (AIOS-RA) for US government and enterprise use, combining Nvidia's Blackwell Ultra GPU systems with Palantir's AIP, Foundr…
Ford Motor Company rehired 350 veteran engineers after automated quality systems and AI tools failed to catch defects in vehicle parts, according to Bloomberg. Kumar Galhotra said the company had over-relied on automated…
AI agents that generate code and orchestrate workflows are becoming production infrastructure, but their non-deterministic outputs create measurement, compliance, and regression challenges. Benchmark systems address thes…
Mozilla President calls for an open-source 'rebel alliance' to counter Big Tech's control over AI, highlighting moves by Canada, the EU, and other nations to prioritize open-source AI infrastructure. Research shows open-…
Canada is prioritizing AI safety and equity at the United Nations, with Ambassador David Lametti dedicating 10 to 15 percent of his time to the issue. Ottawa is collaborating with countries globally to ensure safe and eq…
Veteran investor Jeremy Grantham warned retail savers to avoid US stocks, saying the artificial intelligence boom has created the largest stock market bubble in American history that could end in a 70% decline. Grantham,…
China's Ministry of State Security warned that a popular augmented reality game may be providing geospatial data to train foreign military AI models. The ministry cited reports that an AI company linked to the game harve…
A panel of AI security experts warned that security engineers must evolve from defending deterministic software to protecting probabilistic AI systems, citing threats like prompt injection, data poisoning, and model drif…
Palantir Technologies introduced a new AI engine using NVIDIA Nemotron open models to provide secure, customizable AI for U.S. government agencies in air-gapped environments. The partnership enables agencies to deploy fr…
Mozilla's Zero Day Investigative Network (0DIN) warned that a malicious GitHub repository can compromise a developer's machine via indirect prompt injection targeting AI coding agents like Claude Code, without containing…
An open-source benchmark for prompt-injection detectors has been released, measuring both attack catch rate and false positives on real traffic in a threshold-agnostic way. The benchmark evaluates ten detectors across fo…
Developers implementing ISO 14971 risk management in medical device software must integrate safety considerations throughout the software development lifecycle, from requirements engineering to testing and documentation.…
Google warned that EU competition rules requiring it to share search data with rivals could compromise user privacy, as a senior scientist demonstrated that users can be re-identified from anonymized datasets in under tw…
A developer outlines mobile app security best practices for React Native and Expo apps in 2026, focusing on the OWASP Mobile Top 10. Key recommendations include using expo-secure-store instead of AsyncStorage for tokens,…
Security researchers at Mozilla's 0DIN platform demonstrated that a compromised GitHub repository can infect a developer's machine when an AI coding tool like Claude Code executes its setup, with malicious code loading a…