Sign in Subscribe

Admin

LLMs Show Limited Knowledge of Islamic Law, Study Finds

A new study, 'IslamicLegalBench,' reveals that large language models (LLMs) struggle to provide reliable religious guidance on Islamic law. The best-performing model achieved only 68% correctness with 21% hallucination, and several models fell below 35% correctness while exceeding 55% hallucina

Google Integrates Intrinsic Robotics into Core AI Strategy

Google has integrated Intrinsic, Alphabet's AI robotics project, into its core operations. This strategic move underscores Google's intensified focus on physical AI and robotics, aiming to simplify robot programming and operation. Intrinsic will now operate within Google, collaborating with Goo

Semantic Sensitive Information Poses New Threat in LLMs

Large Language Models (LLMs) are vulnerable to Semantic Sensitive Information (SemSI), where sensitive data can be inferred. Research introduces SemSIEdit, a framework that mitigates this by iteratively rewriting sensitive spans, achieving a 34.6% reduction in leakage with a 9.8% utility loss.

New Defense Thwarts Attacks on AI Language Agents

ICON, a new defense mechanism, effectively neutralizes Indirect Prompt Injection (IPI) attacks on Large Language Model (LLM) agents. By leveraging a probing-to-mitigation framework, ICON achieves a competitive 0.4% Attack Success Rate (ASR) and a significant 50% task utility gain. This ensures

OpenAI COO Hints at Gradual Ad Integration in ChatGPT

OpenAI COO Brad Lightcap announced that the company will introduce ads to ChatGPT users in an iterative process, emphasizing the importance of maintaining user trust and privacy. The rollout began in the U.S. earlier this month for free and Go tier users. Lightcap urged patience as the company

New Defense Mechanism Shields AI Agents from Prompt Injection Attacks

ICON, a new defense mechanism, protects Large Language Model (LLM) agents from Indirect Prompt Injection (IPI) attacks. The probing-to-mitigation framework detects and neutralizes malicious instructions embedded in retrieved content, improving task utility by over 50% and achieving a low 0.4% A

OpenAI, Paradigm Launch EVMbench for AI Smart Contract Audits

OpenAI and Paradigm launched EVMbench on February 18, 2026, a benchmark designed to evaluate AI agents' ability to detect, exploit, and patch high-severity smart contract vulnerabilities. EVMbench aims to advance AI's role in blockchain security and smart contract auditing, potentially setting

AI Framework Boosts Rare Disease Diagnosis from Clinical Notes

RARE-PHENIX, an AI framework, enhances rare disease diagnosis by automating the analysis of clinical notes using large language models. Trained on data from 2,671 patients and validated on 16,357 clinical notes, RARE-PHENIX outperformed existing methods in identifying relevant phenotypes. This

Anthropic Launches Enterprise AI Agents with New Plug-ins

Anthropic has updated its Claude Cowork platform with new plug-ins for finance, engineering, and design, marking a significant push into enterprise AI. The updates automate tasks across various fields and allow Claude to handle multi-step tasks across Excel and PowerPoint, challenging existing

Pentagon Escalates AI Dispute with Anthropic Over Guardrails

The Pentagon has escalated its dispute with Anthropic, demanding the AI company loosen its guardrails by Friday or face penalties. Anthropic refuses to compromise on its ethical stance, particularly regarding mass surveillance and autonomous weapons. The conflict highlights tensions between gov