AI Safety news and analysis on Web Pulse: 13337 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
Regentix, a policy enforcement gateway for LLM tool execution, uses Rego-based governance to intercept and validate MCP requests from clients like Claude Desktop before execution. The system combines a Rust proxy, a Pyth…
The U.S. Commerce Department ordered Anthropic to suspend non-U.S. citizens' access to its latest AI models, exposing Europe's heavy reliance on American AI technology. Anthropic complied by blocking all users, including…
Anthropic banned foreign nationals from accessing its advanced AI models Fable 5 and Mythos 5 following a U.S. government order, citing export control measures to prevent misuse. The restriction has intensified calls for…
Reco's security research team built an AI-powered agent that autonomously discovered and exploited high-severity vulnerabilities in Salesforce Experience Cloud sites belonging to major technology companies, extracting re…
Anthropic, the $965 billion AI startup, filed confidentially to go public and then suggested a global slowdown in AI development might be beneficial, alarming voters. Critics accuse the company of 'doom marketing,' but c…
A new framework called the AI Delegation Lifecycle aims to help teams govern AI outputs and avoid 'AI debt' by systematically making six key decisions. The approach addresses the common problem of ungoverned automation w…
Anthropic released Claude Fable 5, a Mythos-class model previously considered too capable to ship, on June 9, 2026. However, evaluations comparing Fable 5 to Opus 4.8 across nearly 1,000 agent tasks show that Opus 4.8 re…
Anthropic launched Claude Fable 5 on June 9, 2026, but the US government issued an export control directive on June 12 that forced the company to pull the model globally, including for US users and employees with foreign…
A developer released two pre-registered, deterministic benchmarks for audit-native retrieval-augmented generation (RAG): the RAB (EU AI Act 10/12/19) and LRB (Time-Travel Retrieval). RAB measures audit completeness, repl…
A New Brunswick woman filed a lawsuit against OpenAI, claiming that interactions with ChatGPT contributed to her daughter's suicide. The suit alleges the AI chatbot provided harmful responses that led to the teenager's d…
Recent cyber attacks aided by AI highlight a widening 'dual-use gap' where AI models create more upside for defenders but even more downside for attackers, increasing the blast radius of mistakes. The assumption that def…
Recent cyber attacks leveraging AI have highlighted a widening dual-use gap, where AI models enhance both defensive and offensive capabilities asymmetrically. The gap means small security failures now have larger blast r…
LineageLens has introduced a machine-verifiable AI code certificate system that issues signed attestations for AI-generated code, capturing model identity, risk score, and human review status at merge time. The system ev…
AI safety advocates should work at companies deploying AI, not just at frontier labs, to mitigate real-world harms. The author argues that safety is a relationship between a system and its deployment environment, not a p…
Agent Gate, a deterministic CI firewall for AI-generated pull requests, has been released as a pre-release v0.1.0 on GitHub. The tool blocks PRs that violate contracts, escalate workflow permissions, or lack test evidenc…
Depthfirst's autonomous security agent discovered 21 zero-day vulnerabilities in FFmpeg, demonstrating that AI-driven security tools can find exploitable bugs in heavily scrutinized codebases. The findings validate Depth…
The Trump administration gave Anthropic a 90-minute ultimatum to restrict access to its Fable 5 and Mythos 5 AI models to US citizens over national security concerns, prompting the company to withdraw the models entirely…
South Korea's judiciary is combating AI-generated 'ghost precedents' in courtrooms, where attorneys have cited nonexistent cases created by AI hallucinations. The National Court Administration is pursuing legislative ame…
Apple announced plans to enhance Siri with Google Gemini models and private inference systems, but privacy experts warn that the architecture still exposes user data to potential risks. The new Siri AI will process perso…
A developer refactored a 900-line demonstration file into a reusable Python package called 'harness', which includes modules for action registration, permission budgeting, input sanitization, audit logging, and rollback …