Google AI AI News, Models and Product Updates

Google Ai news, model releases, product launches, research updates, and major announcements in one place.

Latest Articles

• Apple brings a fully revamped Siri built on Google's Gemini, but not to the EU — Apple shipped its rebuilt Siri as a beta on iOS 27, running on Google's Gemini models via on-device and Private Cloud Compute; the assistant handles multi-step tasks and screen context for the first time, but EU and China users are excluded due to regulatory constraints.
• Exploring Automated Vulnerability Identification in JavaScript Code Using Large Language Models — Fine-tuned LLMs dramatically outperform SAST tools at detecting JavaScript vulnerabilities: fine-tuned Gemini 1.5 Flash reaches 60% accuracy (up from 29%) versus near-zero for rule-based analyzers, with SQL injection detection hitting 84% in an empirical study of 1,125 snippets.
• ATTRICITE: Training an Open 4B Model for Citation Recovery toward Faithful Attribution — ATTRICITE, an open 4B-parameter model fine-tuned via GRPO on Qwen3-4B, improves scientific citation recovery accuracy from 49.4% to 59.8% on held-out 2025 CS papers — outperforming gpt-oss-20b and landing within 3.9 points of GPT-5.4-mini at a fraction of the parameter count.
• ClinAgent: A ReAct-Based Agent for Conversational Access to Clinical Trial Information — ClinAgent, accepted at CIBB 2026, lets clinicians query clinical trial registries in plain language using a ReAct-based LLM agent — with Gemini 3.0 Flash achieving the highest overall performance and DeepSeek V3.2 thinking mode excelling at planning quality.
• Online Language Adaptive Sampling for Better Distributed Cross-lingual Gains — A trainable adaptive sampling strategy for multilingual model realignment assigns higher sampling probability to languages contributing more to the realignment loss — yielding +0.67 average performance on XLM-R and +0.60 on Gemma 2 9B across tasks. Accepted at EMNLP 2026 Findings.
• AcquireBound: Runtime Authorization for Resources Acquired by AI Agents — AcquireBound is a runtime authorization architecture that quarantines all resources acquired by AI agents — compute, credentials, OAuth tokens, other agents — and activates them only after provenance verification, proving 8 safety properties and rejecting 40/40 unsafe traces in empirical tests.
• Converting Sequenced Fuzzy Cognitive Maps to Causal Virtual Worlds with Large Video Generators — Researchers demonstrate using fuzzy cognitive maps (FCMs) to control causal structure in AI-generated video: FCM dynamics define scene sequencing, a Gemini 3.1 agent writes scripts from FCM meta-rules, and Veo 3.1 renders each scene — shown on a dolphin-shark ecosystem example.
• Stellar Colosseum: A Many-Agent Harness for Long-Horizon Research in Mathematics and Theoretical Computer Science — Google's Stellar Colosseum multi-agent harness achieves 71% on TCS-Bench (research-level theorem proving from FOCS/STOC/SODA papers) and solves 218 of 222 Codeforces problems using Gemini 3.1 Pro — the strongest published result on long-horizon mathematical AI reasoning.
• Hindsight Bias in Clinical Temporal Reasoning: How Future Data Exposure Affects Large Language Model Judgment — A new benchmark of 171 clinical case reports shows LLMs including GPT, Gemma, and Opus consistently fall into hindsight traps when given full patient timelines — rating past decisions with knowledge of outcomes they couldn't have had at the time — and temporal record masking measurably reduces this bias.
• With iOS 27, I’m actually using Siri again — Apple's iOS 27 ships a rebuilt Siri powered by Google's Gemini models, delivering multi-step task execution, on-screen context awareness, and camera-based queries — a genuine upgrade from the limited assistant Siri had been for most users since its 2011 launch.