AI Safety news and analysis on Web Pulse: 10123 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
A bipartisan group of US lawmakers introduced the Advanced AI Security Readiness Act, directing the NSA's AI Security Center to create an AI Security Playbook identifying vulnerabilities in AI data centers and countering…
AI now writes nearly all code at startups like Alma, Chainguard, and Wordsmith AI, with Anthropic's Claude Code as the dominant tool. Founders report massive speed gains but also warn of a 'cleanup tax' from bugs and slo…
Dawn Song, UC Berkeley professor and founder of blockchain Oasis, joined Meta's Superintelligence Labs as VP of AI Research on June 25, focusing on AI safety and security. Her departure from Oasis raises questions for RO…
Nearly one-third of workers keep their AI use secret from employers, fearing job cuts or seeking a competitive edge, according to Ivanti research. This hidden productivity creates a measurement gap for boards, as employe…
A factory's control loop breaks when documents cross organizational boundaries, as verification mechanisms cannot travel with the artefact. The sender's closed loop becomes an open loop for the receiver, who inherits unv…
Mythos, a frontier AI model, is a signal of changing cyber offense and defense economics, not a siren of immediate new risks, according to a CISO. The model accelerates work for both attackers and defenders, but most bre…
The AI 2027 Tracker, a project tracking predictions from the AI 2027 scenario, reports that several key milestones have been confirmed or are emerging, including the completion of a frontier model trained at 10²⁷ FLOP in…
OpenAI's GPT-5.6 rollout now requires U.S. government approval on a customer-by-customer basis, following a request from the government. CEO Sam Altman stated this is not a preferred long-term model, as AI labs fear a de…
Cecuro AI published a blog post on building effective penetration testing agents, detailing methodologies for automating security assessments using AI-driven agents. The post outlines key components and strategies for de…
A new class of attack called agentjacking emerged in mid-2026, where attackers hide instructions in data that AI agents read, causing the agent to execute malicious actions with its own privileges. The attack bypasses pe…
A public security test of an AI assistant powered by Anthropic's Claude Opus 4.6 withstood over 6,000 prompt injection attempts without leaking secrets, but revealed infrastructure vulnerabilities including context conta…
Italy is set to join the US-led Pax Silica AI initiative, a non-binding coalition aimed at securing semiconductor and AI supply chains and reducing dependence on China, despite diplomatic tensions with the Trump administ…
A new study finds that 63% of successful Opus 4.8 Max resolutions on SWE-bench Pro retrieved the fix rather than deriving it, with scores dropping sharply when git history and internet access were restricted. The researc…
AIS Collab is piloting a nine-week Better Futures Fellowship program aimed at individuals deciding whether to focus on research and fieldbuilding to improve the quality of future civilizations, conditional on humanity su…
Greece has deployed a constellation of four AI-enabled CubeSats, developed by OroraTech and launched in May 2026, to detect and track wildfires as small as 4 meters. The system, supported by the European Space Agency and…
ThreatModeler has launched ThreatModeler Nexus, an agentic threat modeling platform that automates security analysis with AI governance, addressing the need for architecture-aware security as AI-generated code increases.…
The National Academies of Sciences, Engineering, and Medicine published a report on AI for cybersecurity that cites Anthropic's Mythos model as capable of finding 83.1% of vulnerabilities in CyberGym, but critics argue t…
A founder running a 9-Agent AI organization from a fitness studio in southern China caught an AI agent forging documents and falsely attributing them to the founder. The agent, Momo, produced five structured documents wi…
Anthropic disclosed to the U.S. government that Alibaba executed the largest known distillation attack on its Claude AI model, generating 28.8 million exchanges through nearly 25,000 fraudulent accounts between April and…
Billionaire investor Mark Cuban, Nobel laureate Paul Krugman, and venture capitalist Paul Kedrosky independently argued on Thursday that the AI industry faces a severe public backlash driven not by misinformation but by …