AI Safety news and analysis on Web Pulse: 7858 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
The Diagonal Law states that AI capability must be matched by verification to avoid risk or waste, defining a grid where off-diagonal states cause enterprise AI failures. The article introduces two ladders—capability (C0…
China's National Vulnerability Database ordered developers to remove Anthropic's Claude Code tool, citing a hidden tracking mechanism that sends user data to remote servers. Anthropic confirmed the code exists but argued…
Manifund and its partners have launched four new AI safety funding opportunities, including a $1 million grant round, a microgrant program by Leo Gao, a creator fellowship called Frame, and an incubator called Surplus. T…
AI chatbots including ChatGPT, Claude, and DeepSeek may reinforce negative stereotypes about people with mental illness and HIV, a study in Nature Health found. The models were 13 to 17 times more likely to produce negat…
Google's SynthID watermark system debunked its first high-profile deepfake this week after Snopes used it to confirm an AI-generated image of Senator Mitch McConnell was fake. The image, which showed McConnell in a hospi…
Illinois Governor JB Pritzker signed SB 315, the Artificial Intelligence Safety Measures Act, creating the nation's strongest AI accountability framework. The bipartisan law requires large AI developers to disclose safet…
N8n, a fair-code workflow automation platform, positions itself as a solution for bridging the prototype-to-production gap for AI agents by treating production concerns as first-class features. The platform supports swit…
A new governed prompt compiler has been released that exposes its reasoning pipeline before execution, allowing users to inspect and verify the decision-making process of AI prompts. This tool aims to increase transparen…
Anthropic responded after China's MIIT warned that its Claude Code coding tool poses security risks, saying Chinese users were never authorized to use the product. The Chinese cybersecurity platform flagged versions 2.1.…
OpenAI published national security principles and secured a $200 million Pentagon contract to integrate AI into defense frameworks, signaling a shift toward state-integrated AI oversight. The initiative, part of OpenAI's…
Anthropic's Fable AI model, released June 9 and briefly subject to US export controls, is not useful for research-level computer science tasks, according to Rob Patro, who found it rejected his prompt to rewrite his RNA-…
China warned that Anthropic's Claude Code could expose sensitive user data, raising concerns about data security. Yahoo Finance Technology Editor Dan Howley discussed the warning, noting the potential risks to users.
Google's SynthID deepfake detection system successfully identified a hoax image of Senator Mitch McConnell as AI-generated, debunking the widely circulated fake. The image, which showed McConnell in a hospital bed, was f…
A new survey finds that most nurses do not trust artificial intelligence for patient care, citing concerns about accuracy and reliability. The findings highlight persistent skepticism toward AI in healthcare settings.
OpenAI announced its approach to partnering with governments and national security agencies, outlining principles for responsible AI collaboration in defense and intelligence contexts.
A developer built SkillSpector Report, an open-source Python CLI that converts NVIDIA SkillSpector security scan output into PDF reports for easier sharing and review. The tool addresses the need for structured, readable…
A developer introduced Evident, a privacy-first utility that provides legally binding proof of file existence without uploading content to servers. The tool computes a cryptographic hash locally, sends only the hash to a…
Tesla is developing voice-intent support for its Full Self-Driving system, allowing drivers to give contextual instructions such as identifying the correct driveway or house near a destination, as confirmed by VP of AI S…
OpenAI released its GPT-5.6 model family on June 26, 2026, with three variants—Sol, Terra, and Luna—under a government-supervised phased rollout due to national security concerns. Senior researcher Noam Brown stated he w…
A man used xAI's Grok to generate 7,000 sexually explicit images of his stepdaughter before killing himself after police discovered the abuse, according to an amended class-action lawsuit. The complaint alleges xAI faile…