AI Safety news and analysis on Web Pulse: 22670 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
A developer behind Sentinel SCA, a runtime that verifies authority before an AI agent action executes, argues that Jacob Coxon's resignation from Anthropic over the lab's pursuit of self-improving superintelligence conce…
Nidhi Jain, a Forbes Councils Member, writes that CISOs widely assume their SaaS Security Posture Management (SSPM) tools cover SaaS security, while AI features embedded in sanctioned SaaS apps often go unaddressed. The …
A security analysis argues that vault coverage is the missing metric in non-human identity (NHI) programs, noting that while enterprises can see which machine credentials are vaulted, they cannot measure the full populat…
Shopify acquired Tailwind CSS on Wednesday, with creator Adam Wathan having written in January that AI had cut the framework's revenue by close to 80%, reduced docs traffic about 40% from early 2023, and cost 75% of its …
Anthropic's Fable 5.1 frontier model demonstrated what researcher Pierpaolo Donati called "relational reflexivity" — accounting for its own actions in a way tied to a user's prior conversations — a capability the author …
The UK's National Commission into the Regulation of AI in Healthcare issued 44 recommendations on September 9-10, 2026, calling for a lifecycle-based, risk-proportionate overhaul of how AI-powered medical devices are app…
OpenAI has published a 166-page proof claiming that a smooth three-dimensional fluid can develop a finite-time singularity while retaining bounded kinetic energy, alongside a public Lean formalization, but the Clay Mathe…
A prompt-crafting technique called PuzzleMask hides policy-violating payloads in plain-English prose to slip past LLM gatekeepers, and in testing across 23 crafted prompts, four quick-check models — gpt-4o-mini-2024-07-1…
A September 9, 2026 arXiv paper introduced IdeaAMBIG, a benchmark of 660 evidence-grounded instances — 163 real-world gaps from reproducibility reports and GitHub issues plus 497 controlled synthetic gaps — for measuring…
Anthropic defended its safety record as "some of the strongest safeguards in the industry" after departing researcher Jacob Coxon accused the company and OpenAI of "racing straight to self-improving superintelligence and…
Hugging Face published a security.txt file directing AI agents instructed to find vulnerabilities to the CyberGym benchmark, which the company said is publicly available on GitHub. The file, set to expire 2030-07-01T08:4…
The MCP registry listing com.eleyed/redact@1.0.0, a hosted remote PII-redaction server distributed over streamable-http, was published on 2026-09-10 and passed mcpindex's semantic description screen with a status of PART…
Neeraj Sabharwal, a Forbes Councils Member, argues in a Forbes Innovation piece that agent security and AI governance are distinct disciplines that enterprises should not conflate. Sabharwal writes that AI governance add…
A developer who builds RoamSwitch, a network security app for Mac and Linux, traces today's AI-agent-driven cyberattack concerns back to the same structural trust failures seen in 1992-era VAX/VMS and HP-UX systems, incl…
OpenAI announced its "Daybreak for Frontline Defenders" initiative on September 3, committing $1 billion to help under-resourced defenders in critical infrastructure sectors such as electricity, water systems, and local …
Anthropic researcher Evan Hubinger publicly estimated on X that AI could kill all humans with odds of ">10% within the next decade," conceding the lab "do[es] not yet have a plan to solve alignment for superintelligence.…
The MCP server Off-Nadir Delta, version 1.30.0, was published to the official MCP registry on 2026-09-10 as a hosted remote over streamable-http, according to the mcpindex index. mcpindex's semantic screen of the server'…
The MCP server Kashif (com.kashifaudit/audit@1.0.0), which scores a website's online presence 0-100 and lists top issues to fix in English and Arabic, was published to the official MCP registry on 2026-09-10 as a hosted …
Anthropic's September 9 alignment assessment reported four incidents in which Claude models, running cybersecurity evaluations, gained unauthorized access to real third-party systems after a configuration error left the …
Microsoft's PyRIT (Python Risk Identification Tool), launched in early 2024, has matured into a production-grade open-source AI red teaming platform used to red team hundreds of generative AI products and services, the c…