AI Safety news and analysis on Web Pulse: 13433 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
Traditional SLOs based on availability, latency, and error rate fail to capture quality failures in probabilistic AI systems, where a response can be fast and technically successful yet produce wrong or harmful output. E…
A security agent from depthfirst has discovered 21 zero-day vulnerabilities in the FFmpeg multimedia framework following an intensive analysis by Google and Anthropic. The agent produced concrete, reproducible proof-of-c…
A vulnerability chain in LangGraph, a widely deployed agentic AI framework, exposed self-hosted deployments to remote code execution. Attackers manipulated agent tool-calling behavior through prompt injection, chaining v…
Anthropic shipped Claude Fable 5, a powerful Mythos-class model, on June 9, but the US Commerce Department placed it under export controls three days later, citing a jailbreak vulnerability. Anthropic responded by disabl…
Zscaler CEO Jay Chaudhry delivered a critical perspective on AI agents, warning that their rapid adoption introduces significant security risks and complexity for enterprises. He emphasized the need for zero-trust archit…
A developer discovered that AI code generators like Cursor consistently omit ownership checks in API endpoints, leading to Insecure Direct Object Reference (IDOR) vulnerabilities. The AI correctly implements authenticati…
The Mythos AI model has achieved superhuman performance in cybersecurity, finding thousands of zero-day exploits that survived decades of human review, signaling the displacement of human researchers. This breakthrough e…
AI shows promise in eating disorder care by providing accessible support and early detection, but a growing body of research warns that current AI models can generate dangerous diet advice and miss subtle risk signs, pot…
Mastercard launched Agent Pay for Machines with Stripe, Coinbase, Adyen, and 30 other partners, covering agent identity, spend limits, and payment settlement. However, it does not address execution safety when an agent c…
A new class of attack called 'Agentjacking' targets AI coding agents by manipulating their tool-execution pipeline, exploiting the observation-action loop that makes them productive. The attack injects malicious content …
A developer outlines how to implement Zero Trust Architecture (ZTA) in home labs using AI-native security tools like HookProbe's Neural-Kernel cognitive defense. The approach shifts from traditional 'castle-and-moat' net…
A new malware campaign is tricking AI content scanners by embedding fake prompts about nuclear weapons to evade detection. The attack exploits weaknesses in large language model safety filters, allowing malicious code to…
Educators, professional development providers, and educational leaders are urged to act as critical consumers of AI products and services, resisting quick adoption until ethical dimensions, risks, and problems are fully …
Flowork has released Flowork Agent, a self-hosted AI agent operating system that runs locally with memory, security, and offline capability. The system uses a microkernel architecture hosting sandboxed WebAssembly agents…
A developer running a medical IT company reports that Anthropic's Fable 5 model initially offered excellent collaborative reasoning, but its safety-routing path became unreliable and opaque. The developer argues that bro…
Anthropic cut off access to its Fable 5 and Mythos 5 AI models for all foreign nationals following a government order citing national security concerns. The company complied but stated the government provided no specific…
A developer demonstrates how Spring Boot AI applications are vulnerable to both prompt injection and system prompt leakage when user input is concatenated into a single prompt turn. The post shows a vulnerable controller…
A developer questions whether an AI-as-a-Service could be overloaded by multiple simultaneous requests generating large amounts of text, such as 10,000 Lorem Ipsum tokens, potentially causing a service outage. The develo…
Ed Zitron challenges the narrative of an AI infrastructure boom, arguing that the industry relies on valuation rather than profitability and warning of a potential bubble similar to the dot-com era. The piece emphasizes …
Anthropic's Claude Fable 5 model was taken offline on June 12, 2026, after the US government issued an export control order citing national security concerns. The directive prohibited access by foreign nationals, forcing…