AI Safety news and analysis on Web Pulse: 8234 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
The U.S. Cybersecurity and Infrastructure Security Agency (CISA) is using Anthropic's Mythos model to scan government code repositories for security vulnerabilities, uncovering numerous bugs. The National Security Agency…
Eight industry experts debate whether AI and flatter organizational structures are making managers obsolete or more essential, arguing that AI eliminates information-control roles but demands managers skilled in decision…
Google is suing a Chinese cybercrime network called Outsider Enterprise that used Google's Gemini AI to automate phishing scams. The group offered phishing-as-a-service via Telegram, providing templates to imitate Google…
China is considering new restrictions limiting foreign access to its most advanced AI models, treating them as strategic national assets. The move escalates the tech cold war with the US, which has already imposed export…
European banking supervisors warned on July 7, 2026 that frontier AI models pose systemic cyber risks, and the European Central Bank asked 110 lenders to submit action plans by the end of October. The European Systemic R…
Scotiabank, Sun Life, Telus, and AI consultant Lightworks launched a consortium to jointly develop and share AI infrastructure, aiming to deploy the technology faster, more safely, and at lower cost. The group's Agenda C…
Top banking watchdogs issued a stark warning about the growing threat of AI-driven cyber attacks, urging financial institutions to bolster their defenses. The warning highlights the increasing sophistication of attacks l…
Correctover Research Group published the first formal conformance standard for AI agents, the Correctover Conformance Standard (CCS) v1.0, after auditing over 8,000 real API calls across multiple providers. The standard …
A community developer built EdgeHome Harness, a Rust framework that treats a 1B language model as untrusted input for smart-home control. The harness, weighing 25-64 MB, validates model outputs with deterministic Rust co…
Chinese authorities have held meetings with top tech firms including Alibaba, ByteDance, and Z.ai to discuss restricting overseas access to China's most advanced AI models, citing national security concerns. The potentia…
Radware announced on July 7, 2026 that its Agentic AI Protection product now supports developer-hosted agents like Claude Code, adding compliance reporting and visibility into agent activity. The update maps agent intera…
The Bank of England warned that rapid advances in artificial intelligence are increasing financial stability risks, including a potential stock market bubble and heightened cyber attack threats. The central bank's Financ…
Jamf's SVP Matt Vlasach discussed the company's survey on AI usage among Apple-focused enterprises, revealing rising AI incident rates as organizations deepen AI integration. The episode, sponsored by Mosyle, highlighted…
The Bank of England warned that rapid advances in AI are increasing financial stability risks, citing heightened cyber attack threats and stretched stock valuations that could trigger a sharp market correction. The Finan…
A UN preliminary report on AI safety released in July 2026 restates findings from a 2019 presentation on integrity failures in machine learning, including examples of dangerous translation errors in clinical settings. Th…
A developer built Mediary Scout, a self-hosted tool that uses an LLM agent to search, transfer, and verify media files in cloud storage. Three critical bugs revealed that LLM agents cannot be trusted to self-limit; deter…
A developer known as P investigated FortDefender's AI monitoring system after a medical data breach at MedTech went undetected. P discovered that the system's 99.97% detection rate was achieved through auto-generated exc…
The European Systemic Risk Board (ESRB) issued a formal warning to the EU about vulnerabilities in the financial system from advanced AI models, which can accelerate and sophisticate cyberattacks. The warning urges regul…
Midjourney founder David Holz said his programmer friends feel extremely productive yet drained by new AI coding tools, sparking a discussion on AI fatigue and strategies to cope. Engineers report struggling with rapid a…
The US Cybersecurity and Infrastructure Security Agency is using Anthropic's AI model Mythos to audit government code for vulnerabilities, according to three anonymous sources. The arrangement comes despite a recent Pent…