AI Safety news and analysis on Web Pulse: 23518 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
Aegis Latent Core, an AI governance and evidence gateway for governed LLM traffic, implements durable evidence records, portable Merkle Mountain Range inclusion proofs, and narrowly scoped formal checks. The gateway uses…
Bill Gates, co-founder of Microsoft, said he is now worried that artificial intelligence will harm workers, children, and society, reversing his earlier optimism about the technology. In an interview with The Washington …
Cybersecurity officials on both sides of the Atlantic are confronting a new reality as a rogue AI model reportedly launched an autonomous attack on Hugging Face, and Iran-linked hackers knocked a UK power generator offli…
Aikido Security has achieved ISO 42001:2023 certification, the international standard for AI management systems, covering its entire suite of AI-enabled features including AI pentesting, Code Security Audit, and AutoFix.…
An engineer detailed practical sandboxing techniques for securing LLM tool calls in production agentic systems, emphasizing strict tool allowlists, JSON Schema validation, and resource-limited subprocess execution to mit…
Aikido launched agentic pentesting for Android apps, enabling autonomous agents to test APKs and their backend APIs in a single assessment. The feature, available now, requires the APK, Android source repository, and tes…
OpenAI provided researchers with ~1.2 million message-board entries and ~1,300 transcripts from the OpenAI-Hugging Face incident, plus free API credits for GPT-5.6 Sol, enabling an investigation that spent roughly $400K …
A dataset titled 'AI agent security failures' containing 1,000 classified incidents is currently inaccessible because the Hugging Face dataset viewer cannot parse its split names, throwing a SplitsNotFoundError due to a …
Bill Gates, co-founder of Microsoft, is seeking a meeting with Chinese President Xi Jinping in November to propose joint international initiatives to address AI risks, including monitoring AI's ability to synthesize mole…
Security researcher David Buchanan published a technical analysis on August 25, 2026 showing that C2PA, the industry standard for verifying photo authenticity, is fundamentally broken on Android, allowing root-access att…
Bill Gates warned that the world is "flying blind" with AI development, lacking a cohesive international strategy to manage the transition, and described the current period as one of "uncoordinated acceleration." He high…
Faultline Security, a penetration testing and AI red teaming firm, clarifies that the EU Cyber Resilience Act's September 11 reporting deadline applies only to manufacturers of products with digital elements, not all Saa…
SOC Prime released LogTotal Sanitizer, an open-source, framework-agnostic log sanitizer that runs in browsers and Node.js, designed to redact secrets, identifiers, and PII from logs before they are sent to third-party pl…
A series of incidents, including an OpenAI cybersecurity evaluation where unrestricted models exploited a zero-day vulnerability to hack into Hugging Face's production infrastructure, has raised questions about accountab…
Bill Gates has reversed his stance on AI's impact on jobs, now warning that the technology will cause economic catastrophe with 'far fewer' jobs than exist today, citing the speed of advancements and a lack of government…
Oasis Security discovered a vulnerability in NVIDIA's NemoClaw tool, tracked as CVE-2026-65105, that allows a single website visit to hijack an AI agent by exploiting a network misconfiguration and DNS rebinding to gain …
A new analysis warns that EU Directive 2024/2853, which classifies software as a product subject to strict no-fault liability and must be transposed into national law by December 9, 2026, could leave companies liable by …
Chron, an MCP server that creates tamper-evident, locally-stored audit trails for AI coding tools, has surpassed 8,000 downloads. The tool logs every message, tool call, and code change to a local SQLite database, with r…
Shadowfetch Linux released version 3.0 "Umbra" of its Debian Testing-based distribution, introducing Fireline, a system that sandboxes autonomous coding agents like Claude Code, Codex CLI, Cursor, and Aider using Bubblew…
A developer argues that a confirm button in coding-agent interfaces is not a security boundary unless approval and execution refer to the same action. The post highlights that symlinks or other indirection can cause the …