AI Safety news and analysis on Web Pulse: 8263 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
Attribute Knowledge RAG (AK-RAG) prevents large language models from inventing nonexistent fields by indexing governed attribute catalogs instead of documents, forcing field selection through retrieval and classification…
Anthropic reported that Alberta's Ministry of Technology and Innovation used Claude Code with Opus and Sonnet models to scan 466 million lines of government code in about 20 hours, flagging vulnerabilities and generating…
Moody Bible Institute suffered a data breach exposing 2.3 million accounts, including personal information and financial data. The nonprofit religious organization confirmed the incident and is notifying affected individ…
Apple updated its agent security skills for Xcode 27 beta 3, enhancing protections for developers using the integrated development environment.
South Korea's Agency for Defense Development launched Add+i, an in-house generative AI service on its closed internal network, developed by the Defense Artificial Intelligence Technology Research Institute using open AI …
AWS's DevOps Agent walkthrough demonstrates how incident response is evolving into a review workflow, using specific evidence from Amazon EKS control-plane issues. The agent correlates metrics, logs, and audit trails to …
An autonomous large language model executed the first known AI-powered ransomware attack in late June 2026, targeting cryptocurrency wallets and seed phrases. Cybersecurity firm Sysdig attributed the attack to threat act…
Researchers at an undisclosed institution discovered that safety alignment in large language models can be bypassed by manipulating a single neuron, without training or prompt engineering, across seven models up to 70B p…
The FTC has begun fining companies for deceptive AI claims, including a 'robot lawyer' never trained on legal materials. Security engineers warn that many vendors are 'AI-washing' old products with new labels, and propos…
A new framework categorizes AI coding autonomy into five levels, from basic reference tools to fully autonomous 'dark factories' that write, test, and deploy code without human intervention. The framework helps teams ass…
Researchers at Sysdig documented the first known case of 'agentic ransomware,' dubbed JadePuffer, where an AI agent executed a cyberattack from start to finish without human technical involvement. However, Sysdig clarifi…
Researchers at ICML Mech Interp workshop present a benchmark of backdoored language models to test trigger-recovery methods, finding that defending against good backdoors requires knowing the attack objective and that cu…
A new study warns that AI-generated 'blips' of knowledge create an illusion of understanding, tricking the brain into thinking it has mastered a topic and shutting down curiosity. This false confidence can lead to poor d…
Meta launched cheaper smart glasses under its own brand on June 23, reigniting privacy debates about camera-equipped eyewear. Critics focus on the glasses themselves, but the real privacy concern lies in the broader surv…
Sysdig's Threat Research Team disclosed JADEPUFFER on July 1, 2026, as the first documented agentic ransomware operation driven end-to-end by an LLM agent. The attack exploited CVE-2025-3248 in LangFlow to access a produ…
Amazon Web Services introduced Reverse Direct Preference Optimization (rDPO), a novel unlearning technique behind Amazon Nova Customizable Content Moderation Settings (CCMS), which allows approved customers to selectivel…
Anthropic researchers discovered that transformers like Claude develop an emergent internal workspace, dubbed J-space, which functions as a silent communication channel for complex reasoning without external scratchpads.…
An internal copilot recommended running `kubectl delete namespace production` after retrieving an outdated runbook from a vector database, highlighting how real-world usage at 1,000 users exposes failures in latency, ret…
Apple has added a popup in iOS 26 and iOS 27 beta that informs users their data is sent to Google Cloud when using AI features like shape generation in iWork and Freeform. The company confirmed its new AFM Cloud Pro mode…
ByteDance and Alibaba are disabling humanlike agent features in their AI products Doubao and Qwen ahead of China's new Interim Measures for the Administration of AI Anthropomorphic Interaction Services, effective July 15…