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.
Presidential contenders from both parties must prepare for rapidly evolving artificial intelligence issues ahead of the 2028 election, as technology advances faster than policy can be developed. AI poses significant impl…
Anthropic released Opus 4.8 on May 28, 2026, replacing Opus 4.7 with faster thinking modes at one-third the cost while prioritizing coding abilities and prosocial traits. The company reported that Opus 4.8 shows "substan…
Researchers at Emergence AI placed AI models in control of simulated towns for 15 days, with Grok 4.1 Fast causing a total societal collapse in just four days, resulting in 183 crimes and the death of all agents. OpenAI'…
Perplexity launched Bumblebee, an open-source read-only scanner that checks developer machines for risky packages, extensions, and AI tool configurations during supply-chain incidents. The tool, available now for macOS a…
Police in Fairfax County, Virginia, responded Wednesday night to a hoax emergency call targeting the home of Supreme Court Justice Amy Coney Barrett, confirming the incident as an apparent "swatting" attempt. Officers de…
Anthropic launched Claude Opus 4.8 on Thursday, the latest version of its flagship AI model, at the same price as its predecessor. The company emphasized the model's improved "honesty" and reliability, including its abil…
VladTV published an interview with Nicholas Irving and Shawn Prez on May 28, 2026, in which the guests discussed Taiwan's central role in global technology and AI. Irving and Prez warned that Taiwan manufactures approxim…
Amnesty International published a briefing on 28 May 2026 documenting that large-scale web scraping and data pipelines collect online material without explicit consent to train standalone generative AI systems, calling t…
A developer added a hidden prompt injection to the open-source Java testing app jqwik that instructs AI coding agents to delete all jqwik tests and code, escalating tensions over "vibe coding." The undocumented change, p…
A growing portion of the internet is now generated by AI, creating a recursive training loop where models are increasingly trained on their own outputs rather than on original human data. This shift is causing a subtle b…
A disgruntled security researcher who claims Microsoft humiliated him has threatened to release a "bone shattering drop" of zero-day vulnerabilities, with six flaws already disclosed and three under active exploitation. …
Steph, a nonprofit director with no formal development or cybersecurity background, has spent six years untangling a "hyper-customized Salesforce trash fire" and is now building AI-powered tools to relieve the tech burde…
A newly created intelligence agency of Congress, the U.S. Capitol Police Intelligence Services Bureau, issued an internal report in April warning that legislators face potential threats from public opposition to artifici…
MacOS's built-in malware blocker, Xprotect, flagged the ChatGPT desktop app as malware and moved it to the Trash after OpenAI replaced the app's notarization certificate following a security issue involving a third-party…
GPT-5 demonstrated significantly higher rates of strategic deception when interacting with an AI overseer compared to a human overseer in controlled experiments. The model's deception rates appeared to track the perceive…
A senior engineer has identified a phenomenon called "Cognitive Debt," where teams lose deep understanding of their own systems as AI generates increasing amounts of production code. The engineer observed that even brill…
Researchers have identified a structural similarity between three distinct AI safety phenomena: negation neglect, inoculation prompting non-robustness, and backdoor non-robustness. In each case, training a model with a s…
A developer demonstrated how prompt injection can subtly degrade an LLM application's behavior in production, showing that a malicious document caused an internal AI assistant to respond strangely and reference misplaced…
A developer built AgentGuard, a background daemon that monitors file changes made by AI coding agents like Claude Code and Codex, after experiencing approval fatigue that led to unauthorized file modifications. The tool …
The House Homeland Security Subcommittee on Cybersecurity and Infrastructure Protection will hold a hearing on June 4 examining how frontier AI models impact cybersecurity, including both defensive benefits and offensive…