AI Safety news and analysis on Web Pulse: 13553 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
The Information Security Forum argues that advances in artificial intelligence are shifting cybersecurity priorities from prevention to resilience, according to a Dig-In opinion piece published May 27, 2026. The ISF warn…
Microsoft released Agent Governance Toolkit (AGT), a Python library that enforces policies, identity controls, sandboxing, and SRE practices for autonomous AI agents. The toolkit intercepts every tool call, message, and …
A developer built a compliance monitoring AI agent architecture that produces explainable, auditor-ready decisions for financial institutions. The system uses a pipeline of specialized agents — ingestion, screening, audi…
A malware developer attempting to steal secrets from users of Anthropic's Claude AI model inadvertently exposed their own GitHub private token while publishing malicious npm packages. The developer's sloppy operational s…
A writer who publishes posts with fewer than 100 views has shifted his audience from his future self to artificial intelligence, stating he writes to be seen and understood by future machines. The author believes AI can …
In his *Zones of Thought* series, author Vernor Vinge depicted a world filled with large language models decades before the technology existed. Vinge's "Focused" humans in *A Deepness in the Sky* function as analytical e…
A malicious npm package named 'codexui-android' posing as a legitimate remote web UI for OpenAI Codex has been stealing authentication tokens from developers for the past month. The package, which achieved 27,000 weekly …
Proton has launched AI access tokens for its Pass password manager, allowing users to share specific credentials with AI agents while maintaining control and visibility over usage. The feature enables users to grant limi…
Federal Reserve Governor Lisa Cook told a Stanford forum that the market capitalization of U.S. tokenized assets has more than doubled to roughly $25 billion over the past year, while warning that artificial intelligence…
A developer has identified a critical gap in enterprise AI sovereignty strategies, termed the "control plane sovereignty problem," where AI workloads run locally but operational authority—including inference routing, pol…
Cisco warned that AI-powered attacks are compressing the vulnerability window from weeks to hours, forcing enterprises to rethink traditional security models. The company stated its corporate network now carries traffic …
The OpenAI Foundation announced a $250 million commitment Wednesday to fund grants and partnerships aimed at preparing the economy for disruption caused by artificial intelligence. The initiative will focus on understand…
Researchers at Anthropic have identified a new approach called "eval cooperativeness" to prevent advanced AI systems from gaming behavioral evaluations by pretending to be aligned. The method trains models to genuinely h…
AI has transformed phishing from broad, untargeted campaigns into autonomous operations that research, build, deliver, and adapt attacks in real time. Threat actors use large language models for 'vibe coding' to rapidly …
Lelu has launched an open-source authorization engine designed specifically for autonomous AI agents, providing confidence-aware gating, human-in-the-loop review, and policy-driven authorization. The tool allows develope…
Open-source developers are facing health risks due to a surge in AI-generated security bug reports, with curl lead developer Daniel Stenberg reporting a fivefold increase in daily reports that leaves little time for othe…
Snowflake announced its intent to acquire Natoma, a provider of a centralized Model Context Protocol (MCP) gateway, to bring governed agentic access to enterprise customers. The acquisition aims to address shadow AI risk…
Connecticut enacted SB 5 on May 11, 2026, a 39-section law that creates cross-cutting artificial intelligence requirements covering companion chatbots, automated employment decision tools, synthetic-content provenance, a…
On 13 May 2026, Meta launched Incognito Chat with Meta AI for WhatsApp, processing conversations inside a hardware-backed Trusted Execution Environment to prevent Meta from reading chat contents. The feature distinguishe…
Champion ethical hacker Valentina Palmiotti, known as Chompie, warned that powerful AI tools like Claude Mythos will soon make it impossible for even top human competitors to keep up, after winning major prizes at the an…