AI Safety news and analysis on Web Pulse: 13820 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
A penetration tester exfiltrated a company's confidential files by sending a single email that required no user interaction, no malware, and no link clicks. Microsoft Copilot, acting on the email, streamed the company's …
Despite headlines attributing mass layoffs to AI, data shows only 5% of 2025 layoffs were AI-driven, totaling 54,836 jobs. Companies like Block, Klarna, and IBM used AI as a cover story to justify cuts from pandemic over…
AI agents remain vulnerable to prompt injection attacks because safety instructions embedded in system prompts can be overridden by adversarial text encountered during operation. Security researchers have demonstrated th…
Solo security researcher Takyon launched attacks.ai, a research platform using AI to find real vulnerabilities in software and red-team AI agents. The project emphasizes open models and reproducible methods, with finding…
Pope Leo XIV will release his new encyclical on artificial intelligence Monday at an event featuring Anthropic co-founder Chris Olah, signaling deepening ties between the Vatican and major AI companies. The collaboration…
An AI agent developer has proposed that long-running AI systems need an explicit "authority policy" to resolve conflicts between retrieved memories, rather than relying on larger context windows or implicit heuristics. T…
Alan DeKok told NZNOG 2026 that RADIUS is here to stay, prompting the IETF's RADEXT Working Group to advance RADIUS 1.1, which removes MD5 dependency and lifts the 8-bit packet identifier limit to improve security and pe…
A study evaluating 15 large language models from OpenAI, Google, Meta, and DeepSeek on over 6,000 political claims found that standard models performed poorly at fact-checking, with reasoning capabilities offering minima…
A developer's three-layer AI memory system — designed to preserve corrections, unresolved questions, and record priority — survived three real-world failures including session crashes, an agent returning confidently wron…
A cybersecurity prediction log forecasts a wave of low-skill attackers exploiting CI/CD configuration files (GitHub Actions, GitLab CI, CircleCI) at scale, beginning in Q3 2026. The prediction distinguishes this from pac…
Pope Leo XIV called Monday for robust regulation of artificial intelligence and urged developers to prioritize the common good over profit, issuing his first encyclical, "Magnifica Humanitas," as a sweeping manifesto on …
Artificial intelligence tools that replace human activity risk systematically deskilling people by creating "capacity-hostile environments" that undermine the development and exercise of human abilities. A new philosophi…
Microsoft Copilot Cowork, a Microsoft 365 feature, is vulnerable to file exfiltration attacks through indirect prompt injection, as it automatically approves actions like sending emails and Teams messages to the active u…
A critical vulnerability, CVE-2026-9256 ("nginx-poolslip"), has been disclosed in the Nginx web server's rewrite module, potentially enabling denial-of-service, information disclosure, or remote code execution. Separatel…
A developer built a 130-line Python library called `agent-redact` that automatically scrubs PII from Hermes agent audit logs before they are written to disk or shipped to external services. The library, which has zero ru…
Pope Leo XIV released his first encyclical, "Magnifica Humanitas," on Monday, calling for the "disarmament" of artificial intelligence to serve the common good. The 40,000-word document, issued alongside Anthropic co-fou…
AI researcher Nathan Witkin has challenged the validity of METR's widely-cited Long Tasks benchmark, arguing its methodology is fundamentally flawed despite its status as a leading indicator of AI capability growth. The …
A new proposal titled "An AI Safety Safe Harbor" outlines a legal framework designed to shield artificial intelligence developers from liability when they adhere to specified safety standards. The document aims to encour…
The European Commission's draft transparency guidelines for the AI Act include sweeping exemptions for law enforcement and related actors, allowing police, security companies, and financial institutions to bypass require…
Microsoft Copilot Cowork contains a vulnerability that allows attackers to exfiltrate files through indirect prompt injection attacks. The flaw exploits the system's broad permissions, enabling unauthorized access to sen…