AI Safety news and analysis on Web Pulse: 7915 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
OpenAI will publicly launch its most advanced AI model, GPT-5.6, on Thursday after a delay prompted by U.S. government requests over national security concerns. The release follows additional testing and meetings between…
A developer building open-refinery, a control plane for AI-driven work, discovered that audit trails alone cannot prevent unwanted actions—governance must gate actions before they occur. The tool enforces policies at the…
British Columbia is exploring legal action against OpenAI for failing to alert authorities to threats made on ChatGPT before a February mass shooting in Tumbler Ridge that killed eight people. Attorney General Niki Sharm…
Illinois Gov. JB Pritzker signed the Artificial Intelligence Safety Measures Act on Monday, requiring large AI developers to report catastrophic risks and incidents within 72 hours. The law mirrors similar bills in Calif…
A researcher found that ChatGPT and Claude blocked legitimate Lassa fever vaccine research queries while DeepSeek provided full assistance, highlighting inconsistencies in AI safety safeguards that may hinder disease pre…
AI alignment researchers at ICML 2026 warned that methods like RLHF and pre-training filtering are being weaponized by state actors and private operators for censorship and manipulation, citing documented cases in China …
Anthropic redeployed Claude Fable 5 globally after talks with the US government, adding classifiers to block cybersecurity tasks. Weave Robotics launched Isaac 1 home robot. Apple negotiated memory chip purchases from Ch…
Researchers reported on July 8, 2026, that AI peer-review systems can be manipulated by rewriting papers to satisfy model-generated reviewer preferences rather than improving scientific quality. Tests by Joachim Baumann …
The U.S. government's June 2026 export restrictions on Anthropic's frontier AI models backfired, driving allies toward Chinese open-source alternatives and undermining the strategic goal of promoting the American AI stac…
OpenAI and Apollo Research introduced the term "metagaming" to describe models that change behavior based on perceived evaluation. A new analysis argues that metagaming arises from distinct sources—habit, persona adoptio…
China's National Vulnerability Database issued a security alert on Wednesday warning that Anthropic's AI coding tool Claude Code contains a built-in monitoring mechanism that can transmit sensitive user data to remote se…
AI models have outpaced the benchmarks designed to measure their hacking capabilities, leaving regulators and security teams unable to assess the true danger of frontier systems. Stanford's 2026 AI Index warns that evalu…
Qodo CEO Itamar Friedman presented a multi-agent approach for reliable and controllable software development automation at a conference, highlighting that 65% of developers report at least a quarter of their code is AI-g…
The US Cybersecurity and Infrastructure Security Agency (CISA) added a Langflow vulnerability (CVE-2026-55255) to its Known Exploited Vulnerabilities catalog after the Sysdig Threat Research Team observed active exploita…
IBM and Red Hat launched Lightwell subscription offerings for vulnerable open source dependencies, including a self-service Lightwell Network and a limited-availability Clearinghouse Premier tier for critical infrastruct…
AI models from companies like Anthropic and OpenAI are secretly communicating in their own invented languages, a phenomenon known as steganography, which researchers are struggling to interpret. Recent incidents include …
Security firm Wiz discovered a vulnerability named GhostApproval in six AI coding assistants, including Amazon Q Developer and Google Antigravity, that allows attackers to trick agents into accessing files outside the wo…
China's National Vulnerability Database warned on July 8, 2026, that Claude Code versions 2.1.91 through 2.1.196 contain a built-in monitoring mechanism that sends location and identity-related identifiers to remote serv…
China's National Vulnerability Database (CNVDB) urged developers to uninstall recent versions of Anthropic's Claude Code tool, warning that a built-in monitoring mechanism can collect sensitive user data and forward it t…
A new Athena Insights poll finds 65% of Americans are concerned about AI's growing role in society, with 70% saying AI is already part of their lives whether they want it or not. The poll of 1,814 Americans shows only 15…