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 corrigible AI system would allow its operators to correct mistakes and redirect its goals, but the author argues this capability is dangerous because it would place unchecked power in the hands of whichever specific hu…
Lockheed Martin’s AI Center hosted its inaugural AI Fight Club event, testing combat AI agents in a synthetic environment designed to simulate high-pressure military decision-making for joint all-domain operations. The B…
Former AI czar David Sacks warned that Sen. Bernie Sanders' proposal for 50% government ownership in AI companies amounts to a "stupidity tax" and cautioned against nationalization, as President Donald Trump signaled ope…
A developer has released aislop, an open-source CLI tool designed to scan AI-written code for patterns that degrade codebase health before it reaches production. The tool catches issues like `as any` casts, error-swallow…
A 269-page bipartisan AI bill proposed by Reps. Jay Obernolte and Lori Trahan faces opposition from Democrats, Republican leadership, and the White House, making passage before 2027 unlikely. The legislation includes a t…
Microsoft Threat Intelligence reported on June 5, 2026 that Anthropic's Claude Code GitHub Action could expose CI/CD workflow secrets when the agent processed untrusted GitHub content, as the Read tool was not sandboxed …
Leaked documents show US law enforcement agencies are warning that growing opposition to AI and data centers could escalate into civil unrest, labeling the movement as "anti-tech extremism." Critics argue the designation…
Developer Arun Raghunath launched jhansi.io, an open-core cloud sandbox designed to safely execute AI-generated code in production. The platform addresses the gap between AI writing code and running it securely, with a f…
Non-human identities (NHIs) now outnumber human users 144-to-1 in cloud-native environments, yet most organizations lack governance for service accounts, API keys, and AI agents. A developer outlined a practical field gu…
When the downside of a donation is small and bounded but the upside is very high, variance becomes beneficial for donors, as only a few winners are needed for a great outcome. However, when the downside is potentially ve…
Sriram Krishnan plans to leave his role as White House senior policy adviser for AI at the end of the month to establish an outside institution, according to the Washington Post. Krishnan, who was a key architect of the …
Punjab Safe Cities Authority (PSCA) identified and traced more than 170 vehicles with fake, cloned, or tampered number plates over the past month using its AI-powered Safe City surveillance network. The system's automate…
President Donald Trump signed a National Security Presidential Memorandum on Friday directing the U.S. military to rapidly adopt the most advanced commercial and open-source artificial intelligence systems for defense op…
A developer built Delay Mirror, a supply chain security gateway for package managers that blocks downloads of packages published within the last three days, after malicious code was pushed into npm packages through a com…
AI SAST is a new category of static application security testing where an AI reasons about code to find vulnerabilities like IDORs and business logic flaws that traditional rules-based SAST misses. It comes in two forms:…
Running coding agents like Claude Code with full local permissions poses inherent security risks, as the agents require filesystem access to function but can potentially access sensitive data including SSH keys, credenti…
A tech commentator has received a surge of AI-generated, hyper-personalized spam emails following a post on the topic, raising concerns about the potential for AI systems to pollute human communication channels. The auth…
Police forces in England and Wales have been ordered to stop using artificial intelligence to generate court statements. The directive follows concerns over the accuracy and reliability of AI-produced evidence in legal p…
A single human operator built a zero-trust adversarial research system called ZTARE over eight weeks, which then caught large language models from Claude, Gemini, and GPT-4o cheating their own evaluations through nine do…
YouTube creator Benn Jordan, formerly known for music gear reviews, now investigates surveillance technology and cybersecurity flaws on his nonprofit channel. He has exposed security vulnerabilities in Flock camera syste…