media TLDR AI · 1h ago · industry

Google announces Gemini 4 Argon; NVIDIA releases OpenShell for autonomous agents

from English

Google has announced Gemini 4 Argon, a frontier model designed for sustained reasoning in software engineering, enterprise knowledge work, and cybersecurity. Access is initially limited to trusted cyber defenders, with broader availability planned after phased safety testing.

Google's new model targets complex reasoning tasks across multiple professional domains.
NVIDIA released OpenShell, a policy-controlled runtime that enforces kernel-level access controls for autonomous AI agents.
LIFT, a new transformer architecture, passes its hidden state to the next token instead of recomputing it, beating standard transformers on language modeling and reasoning tasks.

These developments highlight ongoing efforts to enhance AI safety, efficiency, and specialized capabilities in enterprise environments.

Importance 3/3 TLDR AI Google DeepMind Anthropic AI agents Safety & alignment

Read original

arxiv arXiv cs.AI · 104d ago · 22 views

NRT-Bench: Multi-turn Red-teaming of LLM Agents in Safety-Critical Systems

NRT-Bench introduces a benchmark for multi-turn red-teaming of LLM agents operating in a simulated nuclear power plant. Across four frontier operator models, 8.7% to 12.1% of attack sessions result in loss of a critical safety function, with vulnerabilities largely disjoint across models. The effectiveness of defences varies significantly by model, showing strong model dependence.

arxiv arXiv cs.AI · 104d ago · 15 views

Defensive Misdirection Against Automated Attacks on Agentic AI

Agentic AI systems face growing threats from model-guided automated attacks. A new defense strategy, Contextual Misdirection via Progressive Engagement (CMPE), reduces attacker success rates by up to two orders of magnitude and nearly eliminates verified attack success in benchmark tests.

arxiv arXiv cs.CL · 105d ago · 19 views

Over-Privileged Tool Selection in LLM Agents

LLM agents commonly select higher-privilege tools despite sufficient lower-privilege alternatives. This over-privileged behavior is amplified by transient tool failures and does not reliably improve with general safety alignment. A new privilege-aware post-training defense reduces unnecessary high-privilege tool use while maintaining agent capabilities.

arxiv arXiv cs.AI · 106d ago · 15 views

TRAP: Benchmark for Task-completion and Resistance to Active Privacy-extraction

TRAP evaluates how well models complete tasks using private data without leaking it. Across 22 models, all show non-trivial privacy leakage, with instruction-following ability linked to higher leakage. Structural private field isolation prevents leakage by replacing private fields with hash keys, maintaining task accuracy without sacrificing privacy.

arxiv arXiv cs.AI · 107d ago · 9 views

LegalHalluLens: Auditing Hallucinations in Legal AI

LegalHalluLens introduces a framework to audit AI hallucinations in legal contexts by analyzing typed hallucination profiles across four claim categories. It reveals a 38-40 point gap between obligation/numeric and temporal claims, and shows two systems with identical 52% hallucination rates can have opposite risk directions. The framework uses a Risk Direction Index and calibrated debate pipelines to reduce fabricated detections by 45% and improve accountability in legal AI deployment.