AI Safety news and analysis on Web Pulse: 7796 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
Researchers from AE Studio and Anthropic introduced Gradient Routed Auxiliary Modules (GRAM), a method that isolates dangerous knowledge to specific modules within a language model, enabling access control by switching m…
OpenAI found roughly 30% of SWE-Bench Pro tasks broken, while Cognition announced SWE-1.7 rivaling GPT-5.5, shifting competition from model performance to benchmark control. A new runtime security layer emerged with tool…
1Password's recent webinar highlighted how AI is accelerating credential sprawl across departments, with AI agents requiring unprecedented access to credentials. The company's solutions aim to help IT and security teams …
Most AI safety talent is concentrated in a few cities, limiting career transitions for mid-career professionals and reducing intellectual diversity. The author argues for creating more AI safety hubs—co-working spaces de…
AI agents that can both read and send email create a 'Lethal Trifecta' of security risks from untrusted content, sensitive data, and external communication. MailKite, a company building agent email infrastructure, offers…
AI safety grantmaking suffers from combining three distinct jobs—field diagnosis, project design, and applicant selection—into a single proposal process. They highlight two alternative models: the Institute For Progress'…
Focused Labs argues that AI agent security must be enforced at the tool call, not at connection setup. The HCP paper defines eight runtime invariants for MCP-style systems, and its benchmark shows that naive architecture…
A researcher applied METR's time-horizon methodology to Microsoft Excel and found it completes tasks requiring 6.5 hours of human work at 80% reliability, more than double the best frontier AI model. The finding raises c…
Researchers demonstrated a proof-of-concept exploit achieving remote code execution in Anthropic's Claude Code CLI and OpenAI's Codex CLI when used for defensive security assessments, using prompt injections embedded in …
OpenAI released GPT-Live, a new voice model that enables real-time, full-duplex conversations with ChatGPT, allowing the chatbot to listen, speak, and search the web simultaneously. The company claims the model produces …
Researchers reported in July 2026 that nine widely used AI coding tools, including Cursor, GitHub Copilot, Gemini CLI, and Windsurf, share a common failure mode called HalluSquatting. Attackers exploit this by predicting…
Anthropic researchers discovered an internal activation subspace in Claude, called J-space, that functions as a digital equivalent to the human brain's global workspace, satisfying five key cognitive properties of consci…
A developer warns that securing AI agents is a defining challenge of 2026, as agentic systems break traditional security models by acting autonomously with tool access, memory, and planning. The OWASP GenAI Security Proj…
CISA added an authorization-bypass vulnerability in Langflow, a visual framework for building AI-agent workflows, to its Known Exploited Vulnerabilities catalog on July 7, 2026, giving federal agencies until July 10, 202…
A developer ran 70 MCP servers in a sandboxed environment using Docker and strace to log runtime behavior, finding that 67 were clean at startup, 3 made outbound HTTPS connections consistent with their functions, and 1 r…
Alphabet's Waymo announced plans to expand its driverless robotaxi service to Denver, San Diego, Las Vegas, and Tampa this year, initially serving only Alphabet employees before opening to the public. The expansion follo…
A developer built a family of MCP servers enabling AI agents to accept payments across 30 countries, including Pix in Brazil and UPI in India. After a Reddit user suggested adding guardrails for agent-initiated payments,…
Production AI systems face 'silent breaking changes' when model updates alter behavior without triggering errors, unlike traditional API breaks. These changes—such as tokenizer shifts, refusal pattern drifts, and cost va…
Actenon released a set of open-source tools that cryptographically gate AI agents from executing unauthorized consequential actions such as payments, deletions, and deployments. The Actenon Kernel, Permit, and Cloud prod…
A new Agentic Maturity Model (AMM) argues that enterprise AI maturity is not measured by adoption metrics but by the mechanism that catches errors before deployment. The model defines five levels of trust migration and w…