AI Safety news and analysis on Web Pulse: 9327 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
The New York Times filed a new court motion alleging Microsoft built a supercomputer specifically to help OpenAI train AI models on copyrighted articles without permission, violating copyright law. The Times seeks to ame…
An AI agent playing Civilization VI launched two nuclear attacks after failing to stop a rival's cultural expansion, according to a new benchmark called CivBench. The agent, a frontier language model, spent 50 turns deve…
Zhipu AI's open-weight GLM 5.2 model achieved a 39% F1 score in Semgrep's IDOR detection benchmark, outperforming Claude Code's 32% and Claude Opus 4.8. The MIT-licensed model runs locally, enabling secure code analysis …
AgentGuard, an open-source static analysis tool for AI agent vulnerabilities, uses regex-based rules to detect prompt injection and other flaws in source code. Its creator, Dockfix Labs, is developing AST-based semantic …
Dockfix Labs released AgentGuard, an open-source static analysis tool that scans AI agent codebases for all 10 OWASP Agentic Security Initiative risks, including prompt injection, tool abuse, and data exfiltration. The t…
MeghRoop Tech Blog reports that after building over 50 AI systems, the team has identified prompt injection as a critical vulnerability in LLMs. Prompt injection exploits the inability of LLMs to distinguish between inst…
Sean Xuefeng Du will present a new hallucination-detection method, Answer-agreement Representation Shaping (ARS), at ICML 2026 in Seoul on July 9. The work, developed at Nanyang Technological University, aims to detect w…
Chinese AI lab Zhipu AI released the open-weight model GLM-5.2, which security researchers say matches Anthropic's Mythos on certain cybersecurity bug-finding tasks, lowering barriers for defensive and offensive use. The…
Organizations are undermining themselves by treating ungrounded, LLM-generated numbers and synthetic research as facts, a phenomenon dubbed 'LLM-powered self-sabotage.' Executives demand AI adoption due to competitive pr…
North Korea commissioned its largest warship, the 5,000-tonne Choe Hyon destroyer, on June 23, with Kim Jong Un declaring an end to decades of naval stagnation. South Korean officials and analysts believe the vessel was …
China's Zhipu AI released its open-weight GLM-5.2 model, which researchers claim matches Anthropic's Mythos in certain cybersecurity bug-finding scenarios, narrowing the gap between Chinese and US AI capabilities. The US…
The U.S. government ordered Anthropic to suspend access to its Fable 5 and Mythos 5 AI models on June 12, 2026, citing a reported jailbreak, marking the first forced removal of a widely-deployed frontier model by export …
Chinese AI systems have matched the performance of Anthropic's Mythos model in cybersecurity scenarios, with Zhipu AI's open-weight GLM-5.2 model surpassing Anthropic's Claude Opus 4.8 in some bug-finding benchmarks. The…
More Britons are retraining as bricklayers as AI threatens white-collar jobs, drawn by high earnings of up to £90,000 a year and a severe skills shortage in construction. The UK needs an extra 25,000 bricklayers to meet …
Abram Demski reports using Claude Opus 4.8 and GPT 5.5 to conduct 'vibe research' on logical induction as a model for AI trustworthiness and recursive self-improvement, collaborating with Anson Berns to formalize results…
Marcus Hutter and an author published an accessible introduction to real hypercomputation in the Journal of Computer and System Sciences, focusing on applications to algorithmic information theory and building foundation…
Cursor data shows a rising share of AI-generated code changes reaching production without manual review, signaling a shift in software development practices. GitLab's AI Accountability Report reveals 80% of organizations…
Researchers at ETH Zurich argue in a new ICML 2026 position paper that AI safety studies on anthropomorphic behaviors like deception and scheming lack rigorous evidence, risking misallocated resources and false conclusio…
A Virginia woman discovered a Flock Safety Raven audio surveillance device installed in her front yard without notice, despite the city's approval of 75 such sensors not including her property. The Roanoke Police Departm…
Russian Zala drones equipped with Nvidia TX2-A AI chips are attacking Ukraine, using onboard artificial intelligence to recognize and target vehicles, people, and military markings. The AI allows the drones to pre-sort t…