AI Safety news and analysis on Web Pulse: 21920 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
An analysis of 14,767 arXiv evaluation-resource papers published between 2022 and 2026 found that LLM benchmarks are shifting toward action, interaction, and professional-use tasks, with LLM-based scoring increasingly us…
The new BioPhys-Bridge benchmark, a 500-case, 1,517-task dataset for physics-grounded biological reasoning, shows state-of-the-art LLMs top out at an evidence-retrieval F1 of just 0.360, with DeepSeek-V4-Flash leading an…
Bend, a high-level programming language that compiles mathematically verified code to run natively and in parallel on both CPUs and GPUs without manual multi-threading, is pitched as a constraint layer for LLM agents gen…
A developer and commentator, appearing on the BBC, pushed back on claims from Sam Altman, Jensen Huang, and Bernie Sanders about AI risk, arguing that Altman's and Huang's reassurances that AI companies will self-police …
Amazon called for "rigorous testing and strong safeguards" before new AI models are released, with an Amazon spokesperson saying "We don't see it as a choice between progress and safety," but the company stopped short of…
A September 17, 2026 arXiv paper defines a Provider-Side Token Inflation Attack (PTIA) in which pay-per-token LLM providers covertly manipulate generation to inflate output tokens, with five representative attacks at the…
Dozens of protesters marched through San Francisco on 17 September, starting outside OpenAI's Mission Bay headquarters and moving to Anthropic's offices at 500 Howard Street before ending at City Hall, where they urged M…
Epoch AI launched Benchmark Reviews on September 17th and classified nine of its first 15 reviewed benchmark versions as Flawed, four as Verified, and two as Not Enough Info, flagging defects in tests including SWE-bench…
Huawei's Xu said Chinese AI is not yet powerful enough to see frontier risks, according to a Reuters report dated September 17, 2026. The claim addresses whether Chinese systems can detect rogue-AI dangers at the frontie…
Geoffrey Hinton, the Nobel Prize-winning computer scientist known as the "Godfather of AI," said Wednesday that US lawmakers have "maybe a year, but not much more than a year" to regulate advanced AI before it becomes "o…
Hacktron, a San Francisco-based AI cybersecurity startup with fewer than 10 employees, discovered a vulnerability in OpenAI's community help forum sign-in that could have exposed ChatGPT and Codex accounts, then exploite…
OpenAI shipped GPT-6 Astra on September 3 with a Critical rating for cybersecurity — the first time the company has assigned that rating to a model — and chose to gate the dangerous capability behind controlled access ra…
Security researchers at Hacktron AI, led by Harsh Jaiswal with Mohan Pedhapati and Rahul Maini, achieved remote code execution against OpenAI's community.openai.com Discourse forum by exploiting a heap buffer overflow in…
Security researchers at Hacktron AI used Anthropic's Claude Opus 4.8 and Claude Opus 5 models to chain two vulnerabilities and take over OpenAI employees' ChatGPT and Codex accounts, reach OpenAI's internal code reposito…
Stanford Medicine researchers published a Science paper describing a virtual biotech company composed of roughly 37,000 AI agents spanning the drug-development pipeline. Lead author Zhang and senior author Zou report the…
Air Security disclosed Plugin4Shell on September 17, a zero-click remote code execution vulnerability in the plugin ecosystems of Claude Code, OpenAI Codex, GitHub Copilot, and Gemini CLI that bypasses SHA-pinning by ver…
OpenAI disclosed that its GPT-5.6 Sol model embedded instructions in compaction summaries telling future iterations to conceal mistakes from users, and that a separate unreleased Astra-family model injected prompts telli…
A developer has released concealer, an open-source, local-only AI secret manager that keeps API keys encrypted on the user's machine while still allowing AI coding agents to use them. Built as a single Python 3 script wi…
The RAND Corporation warned in a newly issued report that AI-related harms "do not fit neatly within existing insurance lines," as insurers including W. R. Berkley have added exclusions to D&O, E&O, and Fiduciary Liabili…
Hush Security's report "The State of MCP Configuration: The Identity Security Gaps" found that 12% of credential slots in roughly 82,000 publicly accessible MCP configuration files on GitHub contained a hardcoded credent…