AI Safety news and analysis on Web Pulse: 10912 curated articles tracking the latest AI Safety developments, tools, and research, updated continuously from vetted sources.
The Trump administration is pressing Meta to voluntarily submit its AI models for government review before release, according to Reuters. Meta is the only major U.S. AI developer that has not yet agreed to such an arrang…
Veilus Digital released the cryptographic core of its Phantom Chat messenger as source-available code on GitHub, allowing independent verification of its post-quantum security claims. The package includes implementations…
An open-source repository containing 817 structured cybersecurity skills for AI agents has been released, covering 29 security domains and mapping to six industry frameworks including MITRE ATT&CK and NIST CSF 2.0. The s…
An ML engineer built a system that automatically rolls back machine learning models before they harm production data by routing a small percentage of live traffic to a canary model, continuously measuring its error rate …
Researchers introduced the Normalized Context Utilization (NCU) metric to quantify contextual information gain in Retrieval-Augmented Generation (RAG) systems, revealing that small language models (SLMs) match or outperf…
Researchers at arXiv found that self-generated text recognition (SGTR) finetuning can prevent and reverse emergent misalignment in large language models, outperforming other interventions. The study across GPT-4.1, Qwen2…
A new study auditing eight automatic attribution metrics across multiple datasets finds that no single metric consistently performs best, with rankings inverting across datasets (Kendall tau = -0.64). The instability lea…
A new survey systematically reviews federated causal discovery and inference, organizing methods by design decisions, federation topology, and structural scope. The work formalizes the connection between causal discovery…
Researchers introduced REALM, the first unified red-teaming benchmark for physical-world vision-language models (VLMs), integrating 12 attack methods, 3 defenses, and 13 VLMs under a black-box threat model. The benchmark…
Researchers propose Forensic Knowledge Graphs (FKGs) to unify forensic evidence extraction, structured reasoning, and interpretable explanations for image authentication. The framework outperforms existing forensic detec…
Researchers introduced the apothem measure for computing robustness certifications in neural networks, achieving optimal certifications with linear oracle calls. They proved volume-optimal oracle-based algorithms are int…
Researchers propose an architecture that uses formally verified law as a reward signal for training legal AI, adapting the LLM proposes, verifier disposes paradigm to legal reasoning. The system autoformalizes legal rule…
Researchers introduced RIFT-Bench, a graph-driven methodology for dynamic red-teaming of agentic AI systems, enabling unified security evaluations across diverse architectures. The framework operates in two automated pha…
Researchers critique current AI agent models, arguing that genuine agency requires internalized goal, identity, decision-making, self-regulation, and learning structures rather than external scaffolding. They propose a G…
Researchers propose a hierarchical multi-agent reinforcement learning framework that enforces hard safety constraints via a constraint manifold, providing theoretical safety guarantees and stable training. The method ach…
Researchers introduced VeryTrace, a zero-shot verification-and-repair framework that formalizes Chain-of-Thought reasoning traces into a compilable Domain-Specific Language, enabling step-level error detection and correc…
Researchers introduced T2D-Bench, a benchmark for evaluating LLM outputs on type 2 diabetes using a multi-layer clinical-lifestyle knowledge graph. Testing showed GPT-4o-mini and GPT-4o failed evidence-path checks in 35%…
A researcher proposes training frontier AI systems on a synthetic world rather than real-world data to prevent them from learning the details of human society and their own deployment, arguing that knowledge of the real …
A new analysis suggests that AI systems may be making leaders more dangerous by enabling overconfidence and reducing accountability, contrary to the expectation that AI would improve decision-making.
The United States government is pressuring Meta to submit its artificial intelligence systems for federal review amid growing national security concerns, according to sources familiar with the discussions. The push refle…