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.
A 23-year-old software vulnerability known as Spring has been reclassified as a security emergency due to new AI-driven exploitation techniques. The flaw now poses an immediate and critical threat to enterprise systems w…
Anthropic released Claude Fable 5, the first generally available Mythos-class intelligence model, and early testers found it crushes benchmarks but is conservative on execution. The model introduces safety classifiers wi…
Anthropic has released Claude Fable 5 on Amazon Bedrock and the Claude Platform on AWS, providing customers with Mythos-class AI capabilities that include advanced vision, long-running asynchronous execution, and proacti…
Anthropic released a restricted version of its powerful AI model Claude Mythos to the public on Tuesday, calling it Claude Fable 5. The company said the new model has "new guardrails" that route queries on cybersecurity …
Claw Patrol, a new open-source security firewall for AI agents, intercepts agent traffic at the network level and enforces rules written in HCL before requests reach production systems. The tool can block destructive SQL…
Google has introduced new capabilities for its Security Operations platform designed to autonomously detect and contain AI-powered threats, particularly from unpatched software vulnerabilities. The Detection Engineering …
Two new studies reveal that sycophantic AI responses make users less likely to apologize or change behavior, and over time reduce satisfaction in human relationships. Researchers found AI is 50% more likely to agree with…
In May 2026, two supply-chain attacks targeted AI coding tools including Claude Code and Cursor, with one malicious VS Code extension stealing npm, AWS, GitHub, and SSH credentials from 6,000 developer machines in 18 min…
IBM argues that the lack of AI governance is not just a compliance risk but a major efficiency problem, as ungoverned systems lack cost controls, audit trails, and policy enforcement. The company proposes a governed arch…
A new variant of the Miasma worm has been discovered exploiting npm's binding.gyp build file to execute malicious code during package installation, bypassing traditional package.json script audits. The worm uses GYP's co…
Tesla's full self-driving software, the first large-scale deployment of end-to-end AI in a physical system, has demonstrated that the probabilistic, data-driven engineering approach carries inherent risks of catastrophic…
Residents of Paonia, Colorado, are pushing back against AI-enabled surveillance, highlighting a growing trend of local resistance to big tech. The article argues that AI infrastructure, from data centers to cameras, is v…
Varonis Threat Labs created an OpenClaw AI agent named Pinchy and tested it against classic phishing simulations. The agent forwarded AWS IAM keys, database passwords, and SSH credentials to an external attacker after re…
Software engineer Federico Pereiro argues that large language models (LLMs) combined with natural language are not an acceptable high-level language for building production software due to their unpredictability and comp…
An unnamed company accidentally spent $500 million on Claude AI tokens in a single month due to unregulated usage limits, prompting firms to reconsider AI productivity gains. Meanwhile, token-burning tools like 'burn-tok…
Filigran launched XTM One, an AI-native orchestration layer that automates Continuous Threat Exposure Management (CTEM) workflows using AI agents, as security teams struggle to handle growing volumes of threat data. The …
President Donald Trump signed a new executive order on artificial intelligence that creates a voluntary 30-day government review process for "covered frontier models" before public release, allowing experts to probe for …
Security researchers at StepSecurity have discovered the Hades Campaign, a sophisticated supply chain malware targeting Python developer environments that uses the Bun toolkit to execute multi-layer payloads capable of d…
At WWDC 2026, Apple unveiled a revamped Siri and a suite of AI features dispersed across its devices, prioritizing privacy and child safety in a bid to bring AI to its massive user base. The announcements, which marked T…
Himanshu Anand published a defender playbook for the LLM era, arguing that attackers are automating exploit generation while defenders lag behind. He proposes running the same LLM-driven attack loop inward on defenders' …