AI Safety news and analysis on Web Pulse: 7974 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
A new arXiv paper demonstrates that standard importance weighting fails to correct for selection bias in machine learning when selection depends on unobservables correlated with outcomes. The authors adapt Heckman's 1979…
Researchers developed a power-calibrated statistical framework for logit-based LLM watermarking that establishes explicit relationships between hyperparameters, detection power, and distortion, transforming watermark des…
The US imposed export controls on Anthropic's newest models on June 12, citing jailbreak and cybersecurity risks, highlighting the risk of hold-up for Europe's AI strategy. The authors argue that reliance on foreign-owne…
A developer compiled over 30 references on incident response to create a systematic textbook covering the full lifecycle from detection to mitigation, command systems, investigation methods, human factors, organizational…
A new AI-driven vulnerability discovery platform, 0day Rubbish, uses a multi-LLM ensemble to find and disclose critical zero-days. Its first case study reveals a CVSS 9.8 unauthenticated RCE chain in Cisco CUCM 14.0, spa…
A developer argues that most organizations cannot fully catalog their AI systems, creating a security blind spot. The post introduces the AI-BOM (AI Bill of Materials) as a structured record to track models, datasets, de…
The U.S. Department of Commerce has approved a broad launch of OpenAI's GPT-5.6 model, Axios reported, citing a person familiar with the matter. OpenAI expects to release the model widely this week after additional testi…
Arcade and Manufact have launched agent authorization and MCP hosting products in cloud marketplaces, signaling that agent gateways are emerging as a distinct enterprise category. Nutanix also released its Agent Gateway …
Researchers introduced agent data injection attacks (ADI), a new category of indirect prompt injection that disguises malicious data as trusted data to trick AI agents into executing unintended actions. The attacks bypas…
Maine librarians are offering services to help patrons disable AI features on their devices, driven by concerns over privacy, misinformation, and corporate data collection. Led by Bangor librarian Hannah Cyrus, the movem…
Addy Osmani, former Google Chrome DevRel now at Anthropic, released agent-skills, a repository of 24 workflow skills designed to prevent AI coding agents from skipping critical engineering steps like writing specs, tests…
An engineer used Claude to audit production code and identified 15 vulnerabilities from the OWASP Top 10, compressing a typical 2-3 week audit into hours. The most common findings include SQL injection and NoSQL injectio…
A research group exploring the historical rise of democracy identifies strong civil society, rule of law, and institutionalized political parties as key factors that raise and sustain democratic levels, while warning tha…
A study from Oxford and Potsdam universities found that AI writing tools can inject political bias into user drafts, altering meanings on sensitive topics like abortion and climate change. The researchers warn that these…
Google DeepMind's mechanistic interpretability team proposed a pragmatic framework for validating interpretability tools using proxy tasks, demonstrating its effectiveness by subtracting an "eval-awareness" vector from C…
Nebula Security published exploit code for GhostLock, a 15-year-old Linux kernel vulnerability (CVE-2026-43499) that allows local privilege escalation and container escape. The flaw, introduced in Linux 2.6.39 and fixed …
Engineers face 'cognitive debt' as AI tools write code they don't fully understand, warned by a 2025 MIT Media Lab preprint. A developer recounts two debugging sessions on the same Android codebase: one where Claude Code…
Lilian Weng, cofounder of Thinky, published a summary of 35 papers on harness engineering for recursive self-improvement (RSI), arguing that harness improvements will remain essential even as core models evolve. The post…
Levels of Self, the developer of the x402 agent ecosystem, is offering free certification for the first three external agent operators under its new Certified Agent Standard. The standard requires five criteria: declared…
Meta launched Muse Image on July 7, an AI image generator that lets users create deepfakes of any public Instagram account without consent, with users opted in by default. The tool pulls photos from public profiles to ge…