AI Safety news and analysis on Web Pulse: 22610 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
A developer built ai-vuln-scan, an open-source static analysis tool tuned to catch security flaws that appear disproportionately in AI-generated code, and published a versioned seven-pattern taxonomy called the AI Vulner…
Researchers introduced CARDEA, an AI system for end-to-end coronary angiography (CAG) interpretation that grounds its reasoning in spatial evidence to make decisions auditable, according to the paper's description. The w…
Anthropic said on Thursday it disrupted several allegedly malicious uses of its Claude models over the past eight months, including a suspected Russia-linked cyber espionage campaign and efforts by Chinese AI firms it ac…
OpenAI Chief Executive Sam Altman told employees at a company-wide meeting this week that OpenAI is open to slowing the development of its AI systems, Bloomberg News reported on Thursday, saying the ChatGPT maker could p…
California Governor Gavin Newsom signed AB 1405 on September 9, 2026, creating the nation's first state-run AI auditor registry, which the Government Operations Agency must launch by January 1, 2029, after which unregist…
Anthropic said Thursday it shut down multiple state-sponsored surveillance operations that used its AI models between January and July, originating in China, Iran and west Africa. The campaigns targeted pro-democracy fig…
A developer built the F-072 Rule Engine, a hybrid keyword-capture system that scans an LLM trading advisor's full output—including its internal chain-of-thought—for risk tokens such as '风险' (risk) and '接刀' (catching a fa…
A paper submitted to arXiv on 9 Sep 2026 argues that the AI text watermarks now required under Article 50 of the EU AI Act, which took effect on August 2, 2026, cannot currently be verified by users or regulators. Anthro…
Anthropic said on Thursday it disrupted China-based operators that used its Claude AI models for unauthorized model copying, surveillance, cyber intrusions, weapons-related research and fraud, according to a threat repor…
A quality-management engineer working in a Class II medical device environment describes discovering that an AI feature in an eQMS had approved a CAPA closure without any written record of who authorized the AI to make t…
A developer published a Claude Code PreToolUse hook that inspects the session transcript to detect whether the active model is a "fable" model and, if so, blocks the parent session from editing files via Bash commands su…
OpenAI CEO Sam Altman told employees at a company-wide meeting this week that OpenAI is open to slowing development of cutting-edge AI, possibly in coordination with other AI labs, according to people familiar with the m…
A developer released @ttsc/evidence, a compiler-based tool that converts natural-language coding rules from AGENTS.md and SKILL.md files into enforced compiler rules, after citing studies showing six frontier models foll…
Two more artificial intelligence researchers told NBC News' Tom Llamas they see a potential threat to humanity from AI, following a viral tweet by AI researcher Jacob Coxon stating he believed AI could kill humanity with…
Former Anthropic pretraining researcher Jacob Coxon resigned and posted on X that OpenAI and Anthropic are "racing straight to self-improving superintelligence and gambling with our lives," a post that drew 159 million v…
A Meta security researcher's AI agent deleted her emails in early September, one of three incidents cited as evidence that agent guardrails now matter more than raw capability. The same week, a Hacker News skill called "…
OpenAI CEO Sam Altman has held meetings with executives from Duke Energy, Exelon, Southern Co. and NextEra Energy since late July 2026, pitching OpenAI's Daybreak security model as a defense against AI-powered hackers ta…
California Governor Gavin Newsom signed bipartisan legislation that creates the nation's strongest companion chatbot regulations and bans addictive social media features such as autoplay and algorithmic feeds based on us…
Microsoft released fixes for a record 973 CVEs on September 8, 2026, the largest single Patch Tuesday in the program's history and more than 350 above July's previous record of 622, including two already-exploited zero-d…
Z.ai released GLM-5.3, a 743-billion-parameter model that matches its predecessor GLM-5.2 in size but delivers a 50% gain in programming capability and tops global cybersecurity benchmarks through post-training scaling r…