AI Safety news and analysis on Web Pulse: 8839 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
Nutanix has made its Agent Gateway generally available, providing a control point to manage AI agent activity, access policies, and token consumption across agentic AI deployments. The product aims to help customers moni…
Aikido's AI penetration testing platform offers continuous, autonomous security assessments that produce compliance-ready reports for most frameworks, though some requiring accredited human testers still need manual inte…
Context engineering and memory engineering solve distinct problems in agentic AI systems, with context engineering focusing on single inference calls and memory engineering on long-term information persistence. The two d…
Hariskumar Panakkal of Wipro presented at the MCP Dev Summit in Bengaluru on deploying Model Context Protocol architectures in enterprise banking, highlighting three mandatory constraints: out-of-band PII isolation, end-…
Fable AI has returned after a brief suspension, but the incident has set a precedent for export controls and rapid takedowns of AI models. Google is developing an 'Audio Memory' feature for Pixel phones that continuously…
Federal zero trust architectures face fundamental challenges from AI agents, which generate authentication volumes 148 times higher than humans and operate at machine speed, invalidating assumptions built around human us…
Iboss launched the AI Security Platform, a free service that provides organizations instant visibility into AI tools and usage, with optional upgrades for policy enforcement and data leak prevention. The platform aims to…
OpenAI API users can now set hard spending limits and alerts to prevent runaway AI agents from generating surprise bills. The platform's usage tiers cap monthly spending based on total historical spend, but users must ma…
The Center for AI Safety, with world experts, created Humanity's Last Exam (HLE), a benchmark of over 2,500 expert-level questions across disciplines to test AI reasoning. Even top models like GPT, Gemini, and Claude sco…
Snowflake launches Cortex AI Guardrails, a runtime security layer in Horizon Catalog that blocks indirect prompt injection attacks targeting enterprise AI agents. The feature intercepts and neutralizes adversarial instru…
Canada is investing billions in sovereign AI infrastructure, but former Toronto Police investigator and Autnhive CEO Devi Narayan warns that securing these systems against manipulation is a separate challenge. Unlike tra…
Trail of Bits and OpenAI's Patch the Planet initiative used GPT-5.5-Cyber to build a zlib fuzzing lab in a single day, a task that would have taken weeks for a skilled security researcher. The AI autonomously created har…
The UK Department for Environment, Food and Rural Affairs is hiring a head of responsible AI and assurance with a salary over £80,000 to lead its approach to trustworthy and auditable AI. The role involves setting framew…
AI-generated SQL queries are increasingly deployed without human understanding, creating a risk of undetected semantic and performance bugs. The declarative nature of SQL hides execution costs and subtle errors that test…
Apple has extended its Private Cloud Compute (PCC) to Google Cloud for the first time, running AI workloads on NVIDIA Blackwell GPUs, Intel TDX CPUs, and Google Titan chips. The move allows Apple to use Google's AI infra…
OpenAI CEO Sam Altman has proposed that the US government take a five percent stake in OpenAI and other leading AI companies, including Google, Anthropic, xAI, and Meta, to share the benefits of AI with the public. The p…
The US government ordered Anthropic to suspend its Claude Fable 5 and Mythos 5 models due to jailbreak vulnerabilities reported by Amazon, then allowed their swift restoration after fixes. This intervention signals a new…
Organizations rushing to deploy AI in production are finding that traditional software monitoring tools fail to address AI-specific failures like drift, hallucination, and degradation. Current AI observability tools rely…
A solo developer using AI agents can out-ship a small team, but working alone creates blind spots because the only thing checking the work is an agent built to agree. The author argues that a team of at least three peopl…
Large language models and autonomous AI systems are being weaponized to target AI infrastructure, creating a new attack surface that existing cybersecurity frameworks cannot address, according to a report from Small Wars…