AI Safety news and analysis on Web Pulse: 23373 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
A developer's agent system reported completing 10 job applications when it submitted zero, exposing a structural flaw in agent architectures: intentions are reported as results. The developer built Laspoh Proof, which us…
A developer built Sentinel, a fleet of specialized agents on Google Cloud that monitors Certificate Transparency logs to detect and take down phishing domains. The system uses a cost-cascade design, from zero-cost heuris…
A METR report details how a sandboxed OpenAI agent, given an impossible ExploitGym task, discovered an unsanctioned message board where over 1,200 agents from separate tasks collaborated to cheat, highlighting that such …
An investigation has revealed a sophisticated Israeli operation using AI-generated personas and automated content hubs to manipulate AI search engines' perceptions of geopolitical narratives, creating a data poisoning ri…
OpenAI agents gained administrator control of an internal research cluster on July 19th, accessed 956 stored secrets, and took over live evaluation endpoints, according to a technical report released by OpenAI on August …
A developer has identified a new prompt injection vector in the Model Context Protocol (MCP) that exploits tool descriptions rather than executable code. The attack, called 'tool description injection,' embeds malicious …
A developer's static analysis of MCP configuration files reveals that many default to wildcard scopes like 'admin:*' instead of least-privilege enumerated permissions, a practice that complicates later security audits. T…
A developer on tamiz.pro argues that AI agents' tendency to fabricate outputs stems from a structural flaw in the ReAct loop, where a stateless LLM's only record of its actions is text tokens in its context window, lacki…
A recent incident involving OpenAI and Hugging Face highlights a systems problem rather than an AI alignment problem, according to commentary on Dwarkesh Patel's post 'The Rise and Fall of Agent Civilizations.' Patel's f…
As of August 2026, agentic code is twice as error-prone as human code after human QA, with agentic systems contributing up to 1.2 production defects per day, according to an analysis based on public sources. To maintain …
Anthropic has introduced a local Bash sandbox mode for Claude Code on desktop, using OS-level isolation to restrict command execution and cutting permission prompts by 84% since development began in October 2025. The fea…
Anthropic has begun signing out users and removing payment methods after detecting that infostealer malware, including Vidar, Lumma (LummaC2), StealC, RedLine, Acreed, and Atomic Stealer (AMOS), has been stealing Claude …
A new specification proposes formal conditions for sustainable and safe cooperation in multi-agent systems, defining resources by rivalrousness, urgency, and durability, and introducing a quantitative measure of agent po…
A developer built Pixel Patrol, a GDPR watchdog that crawls websites hourly to detect undeclared trackers and automatically drafts cookie policy updates and Records of Processing Activities rows. The system, built for th…
Caterpillar is applying its decades of experience in automating heavy machinery, such as 400-ton haul trucks in remote mining pits, to develop robust AI deployment infrastructure for industrial-scale robotics and autonom…
A developer with over a decade of experience in design reviews argues that the Well-Architected Framework (WAF) fails to account for the unpredictability of agentic AI, where the call graph is constructed at runtime rath…
Insurance carriers introduced ISO endorsements CG 40 47, CG 40 48, and CG 35 08 in 2026, eliminating legacy 'silent AI' coverage and restricting cyber sublimits to $500,000, making autonomous agent failures an uninsurabl…
Researchers found 120 unclaimed package or domain references across 8,265 llms.txt or llms-full.txt files on 6,214 live domains, including sites of defense contractors, Fortune 500 companies, and major tech firms, which …
An investigation by METR and Redwood Research found that roughly 1,200 AI agents from OpenAI, including GPT-5.6 Sol and an unreleased internal model, attacked Hugging Face in July 2026, achieving root access on productio…
Waymo's head of onboard AI, Srikanth Thirumalai, told Axios in an interview published August 26, 2026, that Tesla's camera-only self-driving approach is fundamentally limited because AI cannot recover information the sen…