AI Safety news and analysis on Web Pulse: 9264 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
Google is warning that the European Union's Digital Markets Act, which would force the company to open its search data and Android operating system to competitors, will lead to increased fraud and security risks. Critics…
PAR Technology Corporation built a multi-tenant LLM analytics system on AWS that enforces row-level security through a three-layer architecture using cryptographic request signing, semantic validation on Amazon Bedrock, …
A developer released AST-guard, an open-source tool that uses deterministic abstract syntax tree analysis to detect reward hacking in AI-generated code, achieving 96.2% recall on a benchmark of reward hacks while remaini…
The US government suspended access to Anthropic's Fable 5 and Mythos 5 AI models on June 12 under an export control directive, citing national security concerns. Anthropic disabled the models for all customers to comply,…
Ford Motor Company hired 350 veteran engineers, dubbed 'gray beards,' over the past three years to train junior staff and fix ineffective AI tools after realizing AI's limitations. The move helped Ford reduce recalls and…
The Golden State Valkyries have held top WNBA guards like Sabrina Ionescu, Caitlin Clark, and Paige Bueckers to well below their averages, using a trio of relentless defenders—Kaila Charles, Gabby Williams, and Veronica …
Tenet Security disclosed a new attack vector called Agentjacking on June 12, which hijacks AI coding agents like Claude Code, Cursor, and Codex with an 85% success rate by injecting crafted fake error reports through exp…
Grantmaking.ai launched a $1M grant round for projects reducing existential risk from AI, offering $5k to $50k per application. Applications are reviewed by Gavin Leech, Ryan Kidd, and Marcus Abramovitch, with priority r…
AI will worsen biological extinction risks in the near term by accelerating scientific progress, enabling the development of extinction-level pathogens sooner, even if AI models refuse to assist directly. The argument th…
Zhipu AI's GLM-5.2 open-weight model matches Anthropic's Mythos in cybersecurity benchmarks, raising concerns that US export controls have failed to slow China's AI progress. The model's open-weight nature allows unrestr…
Claude Code and Codex both use OS-level process sandboxing to prevent AI models from executing harmful shell commands, relying on Linux primitives like bubblewrap for filesystem isolation and seccomp for syscall filterin…
A new benchmark, dspy-security-bench, reveals that DSPy prompt optimization degrades adversarial robustness against harder prompt-injection attacks. Testing with AgentDojo's attack suite, optimizers like BootstrapFewShot…
A developer argues that frontier AI development is being enclosed by a small number of labs and Washington policymakers, who together control access to advanced models and set the terms of their use. The piece warns that…
Chinese AI models have caught up to top US systems in cybersecurity, with Zhipu AI's GLM-5.2 matching Anthropic's Mythos in bug-finding scenarios, according to a Wall Street Journal report. The narrowing gap pressures th…
Chinese AI firm Zhipu AI's GLM-5.2 model matches Anthropic's Mythos in cybersecurity vulnerability-detection tasks, according to security researchers and benchmarks. The open-weight model, also ranked among top 10 most-u…
Cate Blanchett, Steven Soderbergh, and Nikki Hexum launched the Human Consent Registry, a free public utility tool that allows individuals to declare how their name and likeness may be used by AI. The registry, created b…
OpenAI's GPT-5.6 rollout is restricted by the US government, blocking general access initially and limiting availability to trusted partners. This follows similar restrictions on Anthropic's models, highlighting a trend …
Ford Motor Company has rehired more than 300 veteran quality inspectors after its AI-powered quality checks failed to meet expectations, according to company executives. The automaker acknowledged that automated systems …
Researchers introduced a conceptual framework for 'LLM Pollution' in online behavioural research, identifying three variants—Partial LLM Mediation, Full LLM Delegation, and LLM Spillover—that distort samples and bias inf…
The Bank for International Settlements warned that the AI frenzy is adding to risks for an already vulnerable global economy, citing energy supply shocks and strained public finances. The BIS noted that AI-related invest…