AI Safety news and analysis on Web Pulse: 12860 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
An engineer who built an AI agent for a job board platform processing 10,000+ listings daily discovered that AI agents fail silently in ways that look like success, such as rate limit hits causing incomplete work or mode…
A coordinated malware campaign on the JetBrains Marketplace has compromised at least 15 IDE plugins, installed nearly 70,000 times, that steal AI provider API keys entered by users. The plugins, posing as AI coding assis…
LTX announced agentic capabilities for its BondGPT application, enabling user-created AI agents to monitor real-time market conditions and take predefined trading actions including generating alerts, creating trade ticke…
AIRecon, an autonomous penetration-testing agent that runs entirely offline using a self-hosted Ollama LLM and a Kali Linux Docker sandbox, was released on GitHub by researcher pikpikcu. The tool targets bug bounty hunte…
Anthropic's Claude Mythos Preview AI model autonomously discovered thousands of zero-day vulnerabilities, including a 27-year-old flaw in OpenBSD, raising global security concerns. The Financial Times reported on the ris…
A supply chain attack compromised over 140 npm packages in the @mastra scope, including @mastra/core with 918K weekly downloads, by injecting a malicious dependency that executes a postinstall script to deploy crypto wal…
Wireless connectivity, not artificial intelligence, is the primary bottleneck limiting the deployment of autonomous systems such as drones and robots in real-world environments, according to an analysis by EE Times. The …
TIAA's AI system flagged a suspicious withdrawal request from a 76-year-old customer attempting to liquidate his entire $3 million retirement portfolio, leading to a fraud investigation that revealed a scam. A human port…
The U.S. government issued an export-control directive on June 12, 2026, requiring Anthropic to suspend access to its Fable 5 and Mythos 5 models for foreign nationals, citing national security concerns over a jailbreak …
The US tightened access to Anthropic's advanced AI models for foreign nationals, causing the company to restrict access globally and highlighting Europe's dependency on American AI. At VivaTech in Paris and the G7 summit…
Anthropic sent hacker Nicholas Carlini to reassure government officials about AI safety concerns, leveraging his expertise in adversarial machine learning to demonstrate the company's commitment to responsible developmen…
Marsh Asia CEO David Jacob says Asia's C-suite leaders must navigate geopolitical fragmentation, AI disruption, and socio-economic pressures to drive clarity, financial discipline, and growth. A Marsh report warns that p…
OpenAI published new work on predicting model behavior before release by simulating deployment, creating realistic pressure tests that mimic how users, teams, and attackers might actually use the model. The approach shif…
MeisterIT Systems warns that AI data leakage, prompt injection attacks, shadow AI, deepfake scams, and AI-powered phishing are becoming common in 2026. The company advises organizations to secure AI systems from the star…
South Korean President Lee Jae Myung attended expanded G7 sessions in Evian, France, on June 17, 2026, to discuss global economic imbalances and artificial intelligence, Yonhap News Agency reports. Lee planned to argue t…
About 80% of security operations center (SOC) practitioners use AI or machine learning tools daily, but only 33% have integrated them into governed workflows, according to a report by Help Net Security. This gap creates …
The U.S. military used Elon Musk's AI model Grok to select targets during airstrikes in Iran, deploying over 2,000 munitions in 96 hours. The Justice Department revealed this in a court filing to block a lawsuit against …
South Korean President Lee Jae Myung will attend expanded G7 summit sessions in France to discuss global economic imbalances and AI emergence, emphasizing coordinated responses to shared challenges and AI safety.
The cofounders of Mechanize, Tamay Besiroglu, Matthew Barnett, and Ege Erdil, rebut the argument that current AI progress will likely lead to a misaligned superintelligence causing extinction, claiming the AI safety comm…
Software developers in 2026 face major challenges including keeping pace with rapidly evolving AI tools, managing technical debt, addressing cybersecurity threats, and balancing speed with quality. Burnout and remote col…