All news with #google tag

🧭 At Google Cloud we examine how multi-agent systems should delegate tasks intelligently, drawing on Google DeepMind’s Intelligent AI Delegation research. The article outlines four principles: contract-first decomposition, cost-aware model routing, strict data minimization and cryptographic verification, and introducing dynamic cognitive friction to avoid blind compliance. These principles aim to improve reliability, security, and cost-efficiency when agents coordinate in enterprise workflows.

🚀 Google has integrated Antigravity into eligible Gemini Enterprise subscriptions, providing built-in administrative, spend, and security controls alongside new IDE extensions for VS Code and other editors. Administrators can manage pooled quotas, granular spend thresholds, overage options, and centralized usage metrics from the Gemini Enterprise console. The offering aligns Antigravity with Google Cloud's security, audit logging, and data privacy protections while supporting Workforce Identity Federation and ADC for seamless developer access.

🎙️ Smashing Security episode 481 features Graham Cluley with guest Jenny Radcliffe discussing recent social engineering incidents and emerging AI risks. They cover a prank call targeting UK PM Andy Burnham, the timing and effectiveness of social engineering, and demonstrations from Black Hat where researchers jailbroke a Unitree Go2 Pro robot dog using Google’s Gemini via kinetic prompt injection. The episode also notes other industry news and sponsor messages.

🧭 BigQuery Graph introduces measures to unify governed metrics with relationship mapping, enabling agents to reason across complex, multi-hop dependencies without ETL. By mapping tables to an in-place property graph and defining MEASURE in the Property Graph DDL, BigQuery resolves graph paths before computing aggregations using GRAPH_EXPAND and AGG. The release includes a visual graph modeler in BigQuery Studio and native Looker integration to keep business metrics at the data layer.

🔒 A new research paper shows a flaw in how OpenAI, Anthropic, and Google carry encrypted reasoning between API calls, enabling recovery of hidden internal reasoning and secrets from session logs. The team demonstrated replay and decoding attacks that recovered API keys, passwords, and other private artifacts from publicly available agent traces. Vendors implemented mitigations and the authors say the main extraction no longer reproduces as of August 2026, but developers are urged to strip opaque reasoning blocks from shared logs.

🔔 Google reports that Chrome's anti-abuse systems blocked over 7 billion unwanted Android notifications per day in Q1 2026. The company says notification abuse has become a vector for scams, malware, phishing, and fraudulent payment requests, prompting a layered "Swiss cheese" defense model. Chrome now auto-revokes notification permissions from inactive or repeatedly abusive sites and allows users to review and restore access via Safety Hub. The browser also limits message rates for disruptive sites and adjusted permission prompts to be less intrusive on Android.

🧭 This announcement explains how Looker’s governed semantic layer now integrates with Gemini Enterprise, enabling analysts to publish conversational agents via the Agent-to-Agent protocol. The integration routes KPI requests to Looker agents that generate deterministic SQL, preserving version-controlled business logic and existing row- and column-level access controls. It uses a zero-risk pass-through architecture with one-time OAuth consent so Gemini does not ingest or persist enterprise data. Features include native interactive charts, interoperability with other agents, and identity-centric authentication to maintain governance while delivering secure, conversational analytics.

🌐 Google announces the Americas Connect expansion with three new subsea cable systems — Alisios, Canoa, and OlaLuz — plus a new branch on Firmina. These routes will link the Dominican Republic, Panama, Chile, Bermuda, and Florida, enhancing capacity and resilience between Latin America, the Caribbean, the U.S. East and West Coasts, and Europe. The program complements prior investments in Curie, Nuvem, and Sol to strengthen regional connectivity.

🔔 Amazon SageMaker JumpStart now includes Black Forest Labs' FLUX.2-small-decoder and Google's gemma-4-12B-it, expanding foundation model options for AWS customers. FLUX.2-small-decoder offers faster image decoding with lower VRAM use, while gemma-4-12B-it provides unified multimodal understanding across text, image, and audio. Customers can deploy these models via the SageMaker console or the SageMaker Python SDK for scalable AI solutions.

🔍 Malachyte applies attention-based neural networks and LLM-inspired sequence modeling to address the retail "cold start" problem, updating user vectors in real time to personalize search and product pages. By streaming every interaction through Managed Service for Apache Kafka into Bigtable and combining multimodal embeddings, the platform refines predictions and privacy-friendly personalization within 100 milliseconds. Built on Google Cloud's AI stack, the solution leverages GKE, GCE, and Cloud Pub/Sub to enable continuous learning across retailers.

🔐 Three research teams disclosed attacks that bypass passkey protections without breaking FIDO cryptography. SpecterOps showed Windows-exposed signatures chained through Microsoft Entra ID to impersonate privileged users. Unit 42 demonstrated methods to recover synced passkey private keys in Chrome's Google Password Manager, and Dirk-jan Mollema showed malware in a signed-in Windows session could use a Windows Hello for Business key without a fresh PIN. Vendors issued patches and mitigations with differing impacts.

🔎 Google and Mandiant attribute a recent wave of data extortion to UNC6671, which uses vishing to trick employees into spoofed login portals and capture credentials and MFA tokens. The group deploys automated scripts to exfiltrate data from cloud and SaaS environments, including Microsoft 365 and Okta, and operates multiple extortion brands. UNC6671 targets employees’ personal devices, spoofs help desk numbers, and registers adversary-controlled MFA devices to maintain persistence.

🧭 This summer Google Cloud offers a free, hands-on training path powered by Gemini Enterprise Agent Ready (GEAR) to help developers and IT leaders move autonomous agents to production. The program includes sequential courses and skill badges covering agent fundamentals, multi-agent orchestration, ADK engineering, memory and state management, human-centered design, and operationalization on Google Cloud. Participants can earn credentials, access labs and prototypes, and join an All Things Agentic Hackathon with prizes to demonstrate real-world skills.

🛡️ Security advisories from Amazon Web Services, Google, and Vercel detail vulnerabilities in agent harnesses that allowed caller-supplied or forged tool-call data to reach execution without a verifying model turn. AWS patched its managed Bedrock AgentCore service, Google fixed ADK for Python in v2.5.0, and Vercel released fixes for its Codex and OpenCode harness packages. Each vendor's path differed in attack surface and required conditions, and mitigations focus on validating and binding tool invocations to authenticated model events.

🔍 Security flaws in Google’s Agent Development Kit (ADK) workflows could let public-facing AI agents trigger higher-privilege automation, researchers at Pillar Security report. Malicious instructions in pull requests or issues induced agents to post commands that started trusted workflows, enabling actions like altering reviews and extracting tokens. Google removed the affected workflows and applied fixes after disclosure.

🧭 Data Commons on Spanner Graph general availability and a preview of the Data Commons Platform simplify linking private enterprise data with Google's extensive public knowledge graphs. The platform consolidates standardized public datasets from over 100 providers into a unified graph with >400 billion observations and leverages Spanner Graph for native GQL support, incremental updates, and consistent snapshots. Organizations can deploy private instances to federate private and public knowledge graphs while retaining data isolation and enabling natural language query workflows.

🛠️ This article explains how the Google Agent Skills project was launched, structured, and governed to encode Google Cloud domain knowledge into agent-readable instructions. It outlines standardized repository layouts, a CI/CD pipeline with linters and link checkers, and continuous evaluations measuring accuracy and efficiency. The piece also describes ownership rules, internal authoring tools, and a parallel DevRel Skills initiative for internal workflows.

🛡️ Google is developing a Chrome security feature to block policy-installed extensions from hijacking the New Tab page or changing the default search engine on unmanaged consumer devices. The change, spotted in Chromium Gerrit, would enable a feature flag by default to prevent extensions forced by local policies from overriding the New Tab or search settings. Chrome would cancel such installations, record the extension ID as blocked, and avoid repeated download attempts, while allowing administrators an escape hatch policy when needed.

🛡️ Google is rolling out a new two-word naming scheme from its Google Threat Intelligence Group (GTIG) to label cyber threat actors, replacing earlier TAG and Mandiant systems. The first word denotes motivation or activity type, while the second denotes the actor’s origin or status, such as "CASTLE" for China or "RELIC" for Russia, with non-state actors ending in "COMET." The move aims to standardize reporting but risks adding more confusion instead of unifying industry taxonomy.

🔒 Google disclosed fixes for 1,072 security bugs across Chrome 149 and 150, and an additional 370 in Chrome 151, including seven critical issues. The company attributes a surge in discoveries to AI-assisted techniques and is shifting to faster release cadences and automated tooling to shorten disclosure and patch windows. Google is also piloting dynamic patching, session-preserving restarts, and moves toward memory-safe languages like Rust to reduce entire classes of C++-origin vulnerabilities.