AI News Archive: May 19, 2026 — Part 18

Sourced from 500+ daily AI sources, scored by relevance.

• BiRD: A Bidirectional Ranking Defense Mechanism for Retrieval Augmented Generation

  The growing adoption of Retrieval-Augmented Generation (RAG) has led to a rise in adversarial attacks. Existing defenses, relying on semantic analysis or voting, face a trade-off between high computational cost and limited robustness under strong poisoning attacks. Their fundamental limitation is th...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.20123v1
• Awakening the Hydra: Stabilizing Multi-Concept Backdoor Injection in Text-to-Image Diffusion Models

  Text-to-image diffusion models are increasingly developed through open-source reuse and repeated downstream fine-tuning, where reused checkpoints are difficult to verify and thus more susceptible to hidden backdoor behaviors. In such ecosystems, a single pretrained model may be sequentially adapted ...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19698v1
• Exposing Functional Fusion: A New Class of Strategic Backdoor in Dynamic Prompt Architectures

  Existing ViT backdoor attacks based on backbone-overwriting full-tuning are computationally expensive and inflict performance degradation. This has forced adversaries towards the Visual Parameter-Efficient Fine-Tuning (PEFT) paradigm, dominated by adapter-based (e.g., LoRA) and prompt-based (e.g., V...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19478v1
• XAI FL-IDS: A Federated Learning and SHAP-Based Explainable Framework for Distributed Intrusion Detection Systems

  An Intrusion Detection System (IDS) is vital in cybersecurity, detecting unauthorized activity across networks. With attacks on network layers increasing, stronger IDSs are needed. Yet most IDSs rely on centralized detection, forcing IoT nodes to ship data to a server, adding overhead and offering n...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19448v1
• Exploring and Developing a Pre-Model Safeguard with Draft Models

  Large Language Model (LLM) alignment remains vulnerable to jailbreak attacks that elicit unsafe responses, motivating pre-model and post-model guards. Pre-model guards audit the safety of prompts before invoking target models. However, relying solely on the prompt often leads to high false-negative ...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19321v1
• Backdooring Masked Diffusion Language Models

  Masked diffusion language models (MDLMs) are emerging as a compelling new paradigm for text generation, but their training-time security remains largely unexplored. Existing backdoor attacks on Gaussian diffusion models or autoregressive language models do not directly apply to MDLMs because MDLMs r...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19262v1
• Detecting and Mitigating Backdoor Attacks in OTA-FL Systems: A Two-Stage Robust Aggregation Scheme

  Over-the-air federated learning (OTA-FL) improves communication efficiency by exploiting the superposition property of wireless channels, but this same property also creates a critical security vulnerability: the parameter server (PS) cannot access individual local updates, making it difficult to id...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19253v1
• Quantum Machine Learning for Cyber-Physical Anomaly Detection in Unmanned Aerial Vehicles: A Leakage-Free Evaluation with Proxy-Audited Feature Sets

  Unmanned aerial vehicles (UAVs) are cyber-physical systems whose attack surface spans networked avionics and on-board sensor fusion: a compromised GPS or battery module can mimic a benign mission segment and evade naive anomaly detectors. We present a leakage-free evaluation of quantum machine learn...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19233v1
• Token by Token, Compromised: Backdoor Vulnerabilities in Unified Autoregressive Models

  Unified autoregressive models (UAMs) are transformer models that generate text as well as image tokens within a single autoregressive pass. Shared parameters and a multimodal vocabulary simplify the training pipeline and facilitate flexible multimodal generation, yet might introduce new vulnerabilit...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19227v1
• Hunting Vulnerability Variants in AI Infra: Measurement and Reference-Driven Detection

  AI infra has become a shared execution layer for model training, deployment, and agent orchestration. Because many projects reimplement similar model-centric workflows, a vulnerability disclosed in one repository can recur as a variant in another repository with a related design. Yet the prevalence ...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.20051v1
• DASM: Domain-Aware Sharpness Minimization for Multi-Domain Voice Stream Steganalysis

  The growing use of information hiding in network streaming media for covert communication poses a significant security threat, necessitating the development of robust detection technologies. However, existing steganalysis methods for network voice streams mostly rely on data distributions in specifi...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19955v1
• Measuring Safety Alignment Effects in Autonomous Security Agents

  Do stock safety-aligned language models and their uncensored or abliterated derivatives behave differently when run as autonomous security agents? Single-turn refusal benchmarks cannot answer this question: security agents must inspect repositories, call tools, and produce vulnerability evidence ins...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19722v1
• SCARA: A Semantics-Constrained Autonomous Remediation Agent for Opaque Industrial Software Vulnerabilities

  Critical-infrastructure operators are increasingly expected to assess and remediate vulnerabilities in deployed industrial software. However, much of this software exists as opaque industrial software (OIS), including stripped firmware, proprietary protocol handlers, and compiled control logic witho...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19668v1
• Inferring Sensitive Attributes from Knowledge Graph Embeddings: Attack and Defense Strategies

  Knowledge Graphs (KGs) are a powerful representation of linked data, offering flexibility, semantic richness, and support for knowledge enrichment and reasoning. They help data owners organize and exploit heterogeneous data to provide insightful services (e.g., recommendations), yet real-world KGs a...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19644v1
• Devilray: A Systematic Adversarial Model Revealing Blind Spots in Fake Base Station Detection

  Fake Base Station (FBS) detection has been a critical focus of cellular security research for over two decades. However, significant financial and regulatory barriers to accessing commercial FBS (C-FBS) devices have limited direct visibility into real-world operations, forcing detection systems to b...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19232v1
• Can LLMs Produce Better Object-Oriented Designs than Human-Involved Development?

  Background: Large Language Models (LLMs) are increasingly used for code generation. However, their ability to generate multi-class projects that require object-oriented design (OOD) remains unclear, especially relative to projects developed with human involvement. Aims: The primary objective of this...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19901v1
• Prior Knowledge or Search? A Study of LLM Agents in Hardware-Aware Code Optimization

  LLM discovery and optimization systems are increasingly applied across domains, implementing a common propose-evaluate-revise loop. Such optimization or discovery progresses via context conditioning on received feedback from an environment. However, as modern LLM agents are increasingly complex in t...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19782v1
• CriterAlign: Criterion-Centric Rationale Alignment for Code Preference Judging

  Pairwise human preference prediction is central to evaluating code-generation systems, where quality often depends on task-specific trade-offs beyond functional correctness. While rubric-based LLM judges improve interpretability by decomposing evaluation into explicit criteria, most existing pipelin...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19665v1
• Characterizing Real-World Bugs in Tile Programs for Automated Bug Detection

  Tile-based programming frameworks are increasingly adopted to write high-performance GPU kernels in domains such as deep learning and scientific computing. While these frameworks enhance productivity and hardware utilization, their multi-stage compilation pipelines introduce distinct code generation...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19652v1
• Provable Fairness Repair for Deep Neural Networks

  Deep neural networks (DNNs) are suffering from ethical issues such as individual discrimination. In response, extensive NN repair techniques have been developed to adjust models and mitigate such undesired behaviors. However, existing fairness repair methods are typically data-centric, which often l...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19549v1
• MOCHA: Multi-Objective Chebyshev Annealing for Agent Skill Optimization

  LLM agents organize behavior through skills - structured natural-language specifications governing how an agent reasons, retrieves, and responds. Unlike monolithic prompts, skills are multi-field artifacts subject to hard platform constraints: description fields are truncated for routing, instructio...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19330v1
• Does Code Cleanliness Affect Coding Agents? A Controlled Minimal-Pair Study

  As autonomous coding agents see rapid adoption, their evaluation has primarily focused on task completion rates holding the target codebase fixed. This leaves a critical question unanswered: does the structural and stylistic quality, or ``cleanliness'' of the underlying code affect an agent's abilit...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.20049v1
• OpenComputer: Verifiable Software Worlds for Computer-Use Agents

  We present OpenComputer, a verifier-grounded framework for constructing verifiable software worlds for computer-use agents. OpenComputer integrates four components: (1) app-specific state verifiers that expose structured inspection endpoints over real applications, (2) a self-evolving verification l...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19769v1
• When to Answer and When to Defer: A Decision Framework for Reliable Code Predictions

  Code language models are increasingly adopted for both understanding and generative tasks. Despite their success, these models frequently produce overconfident incorrect predictions and underconfident correct predictions, undermining their reliability in deployment. Practical deployment demands thre...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19369v1
• On-the-Fly Input Adaptation for Reliable Code Intelligence

  Code language models (CLMs) play a central role in software engineering across both generation and classification tasks. However, these models still exhibit notable mispredictions in real-world applications, even when trained on up-to-date data. Existing solutions address this by retraining the mode...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19365v1
• MuMuTestUp: Mutation-based Multi-Agent Test Case Update

  Modern software systems evolve rapidly under CI/CD practices, where tests are critical for quality. However, substantial code changes often render existing test cases obsolete, causing pipeline disruptions, reduced productivity, and compromised quality. Recent automatic test update approaches levera...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19265v1
• When Web Apps Heal Themselves: A MAPE-K Based Approach to Fault Tolerance and Adaptive Recovery

  Ensuring the reliability and resilience of modern web applications remains a critical challenge due to increasing system complexity and dynamic runtime environments. This study proposes a modular self-healing framework based on the monitor-analyze-plan-execute over a shared knowledge base (MAPE-K) m...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19261v1
• Fast Multichannel NMF with Block-Diagonal Spatial Covariance Matrices for Efficient Blind Source Separation Using Distributed Microphone Arrays

  Distributed microphone arrays composed of multiple subarrays enable blind source separation over a wide spatial area. Directly applying fast multichannel nonnegative matrix factorization (FastMNMF) to all subarrays can exploit observations from all subarrays, but it requires repeated inversions of l...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19388v1
• Divergence Meets Consensus: A Multi-Source Negative Sampling Framework for Sequential Recommendation

  Negative sampling is significant for training sequential recommendation models under implicit feedback. The predominant strategy, self-guided hard negative sampling, selects negatives based on the model's current state but suffers from three limitations: (1) the coupling between sampling and model u...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19651v1
• Understanding Wacky Weights: A Dissection of SPLADE's Learned Term Importance

  Learned sparse retrieval models such as SPLADE combine the effectiveness of neural architectures with the efficiency of inverted indices. As these models assign weights to terms from a fixed vocabulary, interpretability is often touted as a major benefit of these models. However, the emergence of wa...

  📄 ResearchMay 19, 2026http://arxiv.org/abs/2605.19628v1
• The cost of efficiency in flexible neural representations

  Working memory depends on the flexible representation of stimulus information in neural activity, which changes dynamically depending on task. Stimulus transformations are thought to be efficient in use of neural resources and optimal for task performance. However, these transformations are often opaque, and efficiency may conflict with optimal performance. Here we show that in a working memory task requiring selective recall of one of two stimuli based on a context cue, the prefrontal cortex of two male monkeys prioritized efficiency by overwriting information within a shared neural subspace rather than maintaining distinct subspaces for each stimulus. In neural activity and recurrent neural networks such efficiency incurs a cost, in that efficient representations are more prone to errors. Conversely, stimulation of the cholinergic forebrain which improves behavior altered this default mechanism by encoding distinct contexts in higher dimensions. These findings demonstrate a fundament

  📄 ResearchMay 19, 2026https://www.biorxiv.org/content/10.64898/2026.05.18.722885v1?rss=1
• Developing a multi-modal neuroimaging-based BrainAge model across childhood

  BrainAge models hold promise as a clinical biomarker for developmental brain health, especially in childhood when there is the potential for early intervention. To distinguish between normative developmental variance and pathological divergence, BrainAge models should reflect the dynamic and diverse neurodevelopmental processes that occur in distinct developmental windows across childhood. We utilized multi-modal neuroimaging data from three pediatric cohorts covering ages 4 to 13 years (n = 1005, 2126 scans), split into Train and Test datasets. Twelve sex-stratified BrainAge models were built stratified by type and different combinations of neuroimaging features. Model types were 'Full-Span' models covering the full age range, and 'Phase-Specific' models split into early- and late-childhood. We first compared BrainAge estimates in the Test dataset amongst our candidate models, then benchmarked the best-performing model against published pre-trained models and DNA-based biological age

  📄 ResearchMay 19, 2026https://www.biorxiv.org/content/10.64898/2026.05.19.725847v1?rss=1
• ProtmRNA: Cross-Modal Knowledge Transfer from Proteins to Messenger RNA

  Motivation According to the central dogma of molecular biology, messenger RNA (mRNA) sequences are directly translated into amino acid sequences, positioning mRNA as the fundamental intermediary between genetic information and functional proteins. This natural correspondence suggests that mRNA sequence analysis could greatly benefit from the rich evolutionary and functional representations learned by large-scale protein language models. Results ProtmRNA repurposes the pre-trained ESM-2 protein language model for mRNA sequence processing via cross-modal transfer learning. Evaluated on mRNA- and protein-related datasets, along with eight additional benchmarks compiled in this study, ProtmRNA achieves performance comparable or superior to state-of-the-art mRNA language models while using less than half the pre-training computational resources. This work establishes the potential of cross-modal transfer learning between biological sequences by demonstrating that protein-derived knowledge c

  📄 ResearchMay 19, 2026https://www.biorxiv.org/content/10.64898/2026.05.19.726141v1?rss=1
• Real-World Validation of Machine Learning Models for HIV Treatment Adherence Prediction and Care Gap Quantification: A Multi-Country Analysis of 192,732 Clinical Records

  Delayed diagnosis and poor antiretroviral therapy (ART) adherence remain primary drivers of HIV-related morbidity in low-resource settings, yet real-world AI validation at scale is lacking. We conducted a retrospective validation study using two publicly available, de-identified datasets: a Quality of Care cohort of 27,288 HIV-positive patients on ART across multiple healthcare facilities, and the CEPHIA multi-country assay database comprising 165,444 specimen records from six countries. Four machine learning classifiers were evaluated using 10-fold stratified cross-validation with SMOTE applied strictly to training folds. Explicit data leakage prevention, ablation analysis, calibration assessment, and bootstrap confidence intervals were applied. Economic projections used one-way sensitivity analysis. This study adheres to TRIPOD reporting guidelines. Random Forest achieved AUC-ROC of 0.9753 (95% CI: 0.970-0.975), sensitivity 87.3% (95% CI: 86.4-88.2%), specificity 95.7% (95% CI: 95.2-

  📄 ResearchMay 19, 2026https://www.medrxiv.org/content/10.64898/2026.05.15.26353325v1?rss=1
• Satellite imagery encodes features predictive of regional mortality and life expectancy

  Background Increasingly accessible satellite imagery provides scalable measures of the built and natural environment relevant to population health. However, whether such imagery can capture subnational variation in mortality and life expectancy remains unclear. We therefore assessed its predictive value for regional mortality and life expectancy across OECD regions. Methods We conducted an ecological, cross-sectional prediction study using 2023 data from OECD Territorial Level 3 (TL3) regions. Annual cloud-masked composites from the Harmonized Landsat and Sentinel-2 collection were processed in the Google Earth Engine, tiled at 224 x 224 pixels, and encoded with the pretrained Prithvi foundation model to derive region-level satellite embeddings. For each outcome, we trained LightGBM regressors for a country-only baseline, a satellite-only model, a combined model (country + satellite), and a final contextual model that additionally included prespecified socioeconomic and environmental c

  📄 ResearchMay 19, 2026https://www.medrxiv.org/content/10.64898/2026.05.17.26353439v1?rss=1
• Predicting Intensive Care Readmission Among Hospitalized Children

  Objective: Readmissions to the PICU are associated with increased morbidity and mortality. A prediction model that can identify children at risk of readmission at the time of transfer can allow providers to intervene and potentially improve patient outcomes. The objective of this study was to derive and validate machine learning models to predict PICU readmission at the time of transfer. Design: Retrospective observational cohort study Setting: Three quaternary care PICUs in the city of Chicago Patients: All children admitted to the PICU between 2012 and 2019. Measurements: The primary outcome was unplanned readmission to the PICU within 48 hours of transfer to the inpatient ward. Predictor variables included vital signs, patient characteristics, and laboratory results. We developed and externally validated four models to predict PICU readmission: logistic regression, elastic net, random forest, and XGBoost. Main Results: This study included 35,601 patients, with readmission rates rang

  📄 ResearchMay 19, 2026https://www.medrxiv.org/content/10.64898/2026.05.15.26353330v1?rss=1
• Predicting Distant Melanoma Metastasis at Diagnosis Using Machine Learning

  Distant melanoma metastasis at the time of diagnosis is uncommon, but has major implications for patient prognosis and treatment selection. However, few tools can reliably predict the risk of distant metastasis at initial presentation. Here, we developed and evaluated machine learning models to predict distant melanoma metastasis using routinely captured clinicopathologic and demographic variables across all histologic subtypes. Using the National Cancer Institute Surveillance, Epidemiology, and End Results (SEER) program from 2010-2022, we identified adults aged 20 to 90 years with melanoma as the first and only primary malignancy (n=51,285). Explainable Boosting Machine achieved a strong balance of discrimination and precision (AUROC = 0.947, AUPRC = 0.610, Precision = 0.793, Brier = 0.015). At 90% sensitivity, specificity was 0.843 with consistent performance across cross-validation folds. Clinicopathologic variables, including T stage, Breslow thickness, ulceration, and mitotic act

  📄 ResearchMay 19, 2026https://www.medrxiv.org/content/10.64898/2026.05.14.26353271v1?rss=1
• How to use Google’s new AI agents to go beyond your standard searches

  Google is launching AI-powered “information agents” that can monitor topics in the background and proactively alert users to updates and changes.

  🌐 MovesMay 19, 2026https://techcrunch.com/2026/05/19/how-to-use-googles-new-ai-agents-to-go-beyond-your-standard-searches/
• An interpretable and interactive clinical AI agent for personalized anti-infective decision support in carbapenem-resistant Gram-negative bacterial infection

  Carbapenem-resistant Gram-negative bacteria (CRGNB) infections remain difficult to manage because treatment decisions must balance heterogeneous patient risk, limited antibiotic options, potential toxicity and emerging resistance. Clinical care in this setting requires not only single-endpoint risk prediction, but also decision-support frameworks that can jointly enable prognosis assessment, result interpretation, and individualized treatment comparison. Here we present Dr.BUG, an interactive clinical AI agent for personalized decision support in CRGNB infection. Dr.BUG integrates stable feature-set selection, multi-task prognostic modelling, interpretability analysis and model-based simulation of antibiotic regimen recommendation into a unified workflow. Using a development cohort, a temporally independent validation cohort, and external cohorts from the MIMIC-IV dataset, we developed and validated models for four clinically relevant tasks: clinical efficacy, survival outcome, polymyx

  📄 ResearchMay 19, 2026https://www.medrxiv.org/content/10.64898/2026.05.18.26353005v1?rss=1
• With Gemini 3.5 Flash, Google bets its next AI wave on agents, not chatbots

  Google launched Gemini 3.5 Flash, its most powerful coding and agentic AI model yet, at the company's annual developer conference. It is capable of autonomously executing complex tasks and building software from scratch.

  🤖 ModelsMay 19, 2026https://techcrunch.com/2026/05/19/with-gemini-3-5-flash-google-bets-its-next-ai-wave-on-agents-not-chatbots/
• Google updates its Gemini app to take on ChatGPT and Claude at IO 2026

  The updates signal Google’s push to turn its Gemini app into an all-purpose AI hub rather than a stand-alone chatbot.

  🌐 MovesMay 19, 2026https://techcrunch.com/2026/05/19/google-updates-its-gemini-app-to-take-on-chatgpt-and-claude-at-io-2026/
• Google introduces Gemini Spark, a 24/7 agentic assistant with Gmail integration, at IO 2026

  At the Google I/O developer conference, the company announced a new agentic personal assistant called Gemini Spark, built from Gemini's base models and an agentic harness from Google Antigravity.

  🌐 MovesMay 19, 2026https://techcrunch.com/2026/05/19/google-introduces-gemini-spark-a-24-7-agentic-assistant-with-gmail-integration/
• Gemini 3.5 Flash might be fast enough for gen AI to make sense

  Google says its more efficient Gemini 3.5 Flash is the key to your agentic AI future.

  🤖 ModelsMay 19, 2026https://arstechnica.com/google/2026/05/google-announces-agent-optimized-gemini-3-5-flash-and-a-do-anything-model-called-omni/
• Everything Announced at Google I/O 2026: Gemini, Search, Smart Glasses

  Google is sprucing up its Gemini models, revamping search, and enabling AI agents in everything. There are also some spiffy new smart glasses coming this fall.

  🌐 MovesMay 19, 2026https://www.wired.com/story/everything-google-announced-at-google-io-2026/
• Gemini Spark Is Google’s Response to OpenClaw’s 24/7 AI Agent

  Google’s always-running, data-hungry AI agent is designed to spend your money and send your emails.

  🌐 MovesMay 19, 2026https://www.wired.com/story/googles-response-to-openclaws-24-7-ai-agent/
• Google I/O 2026 Live Blog: All the Gemini and Smart Glasses Updates as They Happen

  Follow our live coverage of Google’s annual developer keynote, where the company will announce updates to its Gemini suite of AI tools and more details about Android XR smart glasses.

  🌐 MovesMay 19, 2026https://www.wired.com/live/google-io-2026-live-blog-gemini-android-xr-search/
• Google’s new Gemini Spark is an always-on AI agent for daily digital tasks

  Entering into the artificial intelligence agent era, Google LLC today introduced Gemini Spark, a 24/7 personal AI assistant that can help people navigate their digital lives and do real work on their behalf. Under the hood, Spark runs on the newly released Gemini 3.5 Flash and uses the company’s updated Antigravity platform to orchestrate AI agents. […] The post Google’s new Gemini Spark is an always-on AI agent for daily digital tasks appeared first on SiliconANGLE .

  🌐 MovesMay 19, 2026https://siliconangle.com/2026/05/19/google-unveils-gemini-spark-always-ai-agent-daily-digital-tasks/
• Google’s new $100 AI Ultra plan just changed the AI race — here's what you get

  Google’s new $100 AI Ultra plan just changed the AI race — here's what you get Tom's Guide

  🌐 MovesMay 19, 2026https://www.tomsguide.com/ai/googles-new-usd100-ai-ultra-plan-just-changed-the-ai-race-and-gemini-spark-is-the-biggest-reason-why
• Google unveils Gemini Spark — a '24/7 personal AI agent'

  Google unveils Gemini Spark — a '24/7 personal AI agent' Tom's Guide

  🌐 MovesMay 19, 2026https://www.tomsguide.com/ai/google-gemini/google-unveils-gemini-spark-a-24-7-personal-ai-agent-that-could-be-a-game-changer-for-agentic-ai
• Google’s new AI agent can draft your emails, monitor your inbox and eventually spend your money

  Google on Tuesday unveiled Gemini Spark , a personal AI agent designed to work around the clock — drafting emails, assembling documents, monitoring inboxes, and eventually making purchases — even when a user's laptop is closed and their phone is locked. The announcement, made at Google I/O 2026 , is the company's most ambitious attempt yet to transform its AI assistant from a tool that answers questions into one that autonomously completes tasks. It also arrives at a moment of extraordinary competition, as Microsoft, OpenAI, Anthropic, and Apple all race to build AI systems that don't merely converse but act — completing multi-step workflows with decreasing human supervision. "We are in that part of the cycle where people want to see real value in the products they use on a day-to-day basis," Sundar Pichai, CEO of Google and Alphabet, said during a press briefing ahead of the keynote address. With Spark, he argued, that value comes from an agent that never stops working. It operates ar

  🌐 MovesMay 19, 2026https://venturebeat.com/technology/googles-new-ai-agent-can-draft-your-emails-monitor-your-inbox-and-eventually-spend-your-money