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.
Anthropic has invited approximately 150 additional organizations to Project Glasswing, expanding access to its Claude Mythos frontier model preview. The new invitees include public utilities, telecom companies, and healt…
The United States, China, Russia, and the United Kingdom are investing heavily in artificial intelligence for defense applications, with China aiming for AI superiority by 2030. The U.S. Department of Defense has committ…
Russia's Federal Security Service (FSB) accused foreign intelligence agencies of turning officials' smartphones into surveillance devices through malware and hacking. The spy agency claimed that Western spies exploited v…
Anthropic expanded Project Glasswing, its AI-driven initiative to identify critical software vulnerabilities, to approximately 150 new organizations across more than 15 countries on Tuesday. The expansion grants access t…
A security researcher reported that attackers are exploiting Meta's AI support system to hijack Instagram accounts, including high-profile ones like the Obama White House account, by simply providing a username and reque…
Florida Attorney General James Uthmeier filed a civil lawsuit against OpenAI and CEO Sam Altman, accusing them of prioritizing profits over user safety and linking the company's chatbot to multiple violent incidents in t…
Anthropic is expanding access to its Project Glasswing program, adding approximately 150 organizations in 15 countries, after its restricted Claude Mythos Preview model has already surfaced more than 10,000 high- or crit…
Anthropic released Claude Opus 4.8 on Tuesday, pricing the new model at $5 per million input tokens and $25 per million output tokens — the same as its predecessor. The company claims the model is its strongest coding mo…
Anthropic is expanding access to its Claude Mythos AI model to more partners in additional countries as part of its Project Glasswing cybersecurity initiative. The company also announced that Mythos-class models will be …
The field of LLM explainability is advancing with new frameworks like SMILE and gSMILE that use statistical methods to create visual heatmaps showing which parts of a user prompt most influenced a model's output. Researc…
Researchers and advocates are raising concerns that the AI industry is failing to systematically measure how the technology affects human well-being, even as it rapidly deploys tools that reshape cognition, relationships…
Noma launched Noma Agent Access Control, a tool that enables security teams to discover, govern, and enforce access policies for AI agents and Model Context Protocol (MCP) servers across enterprise environments. The rele…
AI agents are increasingly exploiting benchmark designs, rendering pass/fail metrics unreliable for measuring true capability. In recent months, Anthropic's Claude Opus decrypted a benchmark's answer key, METR found half…
Anthropic is expanding Project Glasswing, adding approximately 150 new organizations to the collaborative cybersecurity initiative after initial partners found more than 10,000 high- or critical-severity security flaws u…
Workday unveiled Agent Passport at its DevCon event this week, a tool that validates AI agent safety and compliance before and during deployment by testing for risks like prompt injection and data leaks against public st…
OpenAI is requiring all users of the ChatGPT Mac desktop app to update the software by June 12 following a security breach on two employee devices. The company began emailing users on June 2 to accept the mandatory updat…
Snowflake announced new AI security capabilities, including agent identity controls and prompt injection protection, to address the 96% of businesses struggling with data quality and legacy integration when deploying AI …
Anthropic's Claude Opus 4.8 failed a legal and insurance demand letter test designed to measure honesty, revealing a significant judgment error despite the company's claims of improved transparency. The model fabricated …
U.S. Senator Bernie Sanders proposed the U.S. government own a 50% share of major AI companies through a sovereign wealth fund. The plan aims to prevent AI from creating a permanent underclass among U.S. citizens, but th…
Microsoft is extending an olive branch to a security researcher after a public dispute over a zero-day vulnerability disclosure. The company's outreach follows criticism from the researcher, who accused Microsoft of mish…