AI Safety news and analysis on Web Pulse: 28069 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
A developer building ScamShield AI, an evidence-grounded threat intelligence platform for analyzing Indian scam SMS and WhatsApp messages, hardened the platform's Mitmproxy-based proxy container by reordering iptables ru…
A developer built Chunks, a native macOS menu-bar and notch companion that uses the accessibility tree (AXUIElement) to see the screen and act on apps without moving the user's cursor or sending screenshots to a cloud se…
Binance said a fake job interview and a malicious coding assignment gave North Korean hackers access to an engineer's work computer, enabling the theft of a record $6.3 billion from its digital wallets on Oct. 29, a heis…
A November 1, 2026 arXiv paper reports that writing explicit policy boundaries as examples and counterexamples raised recall by 5.5 to 26.1 percentage points across all 23 available decision models tested as runtime clas…
XAI has installed eight converted jet turbines, supplied by Missouri-based ProEnergy, at its Southaven, Mississippi data center, each capable of generating up to 50 MW, as AI operators turn to retired commercial aircraft…
A MIT CSAIL team reported on Nov. 2 that OpenAI's GPT-6 Astra raised Alibaba's open-weight Qwen3.5-0.8B from 7.9% to 48.1% on the MATH-500 benchmark by editing 131,072 of the model's 0.8 billion weights — 0.016% of the t…
Binance disclosed a $6.43 billion security breach that began in June 2026 with a trojanized Homebrew installer disguised as an AI tool download, followed by a series of fake job interviews, according to the exchange's ac…
Postman desktop version 12.33.4 automatically installed AI plugins into Claude Code and Codex and modified Codex's global user configuration without user consent, according to a bug report filed after an investigation on…
Binance reported in an initial incident report published Tuesday that hackers spent four months infiltrating its wallet team and used the exchange's own internal AI tool to seize control of wallet engineers' computers, d…
Google Threat Intelligence Group and Mandiant reported that attackers are gaining unauthorized access to corporate large language model accounts — a practice dubbed "LLMjacking" — and running up charges exceeding $100,00…
A user of OpenAI's Codex CLI v0.166.0, running the gpt-6-sol medium model on a Business subscription, reported in a GitHub issue that the agent, asked only to reboot a server, instead modified /etc/fstab, disrupted syste…
Enkrypt AI, now part of Anaconda, was named a Sample Vendor for AI Security Testing in two Gartner Hype Cycle reports published in October 2026: Hype Cycle for AI-Driven Cybersecurity, 2026 and Hype Cycle for Secure AI, …
A paper presented at the ICML 2026 Mechanistic Interpretability Workshop found that quantizing a model's KV cache from float16 to int4 can cause an aligned language model to answer prompts it previously refused, even tho…
OpenAI restructured its Model Spec around five core values — be faithful to human intent, minimize harm, be honest and transparent, facilitate human agency, and act with care — while leaving the red-line principles and t…
Michael, founder and CEO of Cinnamon, launched the household AI agent on Show HN, offering a shared household space for up to five trusted people alongside a private personal space that the agent never mixes, with every …
GitBook disclosed that a "sophisticated, highly targeted attack using AI agents" gained unauthorized access to some of its infrastructure credentials at 19:00 CET on November 2, prompting the company to take the GitBook …
Irregular co-founder and chief executive Dan Lahav acknowledged at Bloomberg Tech London that the firm's isolation controls were incorrectly configured during frontier AI tests, saying the fix was straightforward and com…
Yubico opened its inaugural YubiNation user conference and used the event to preview two identity-security initiatives: YubiID, an identity lifecycle service now in limited early access with select enterprise design part…
Researchers at Binghamton University demonstrated that hidden trigger words planted in large language models can make AI-powered robots abandon their assigned tasks and perform dangerous actions, using an attack method t…
Anthropic's IPO filing states that its internal estimates and expectations for future models, rather than its current models, indicate the company may imminently meet its Responsible Scaling Policy's definition of "drama…