< ciso
brief />
Tag Banner

All news with #google tag

796 articles

Signal Adds Protections to Block Siri and Gemini AI

πŸ”’ Signal has introduced new protections that stop AI assistants such as Siri and Gemini from capturing private conversations on iOS 27 and supported Android devices. The safeguards also block screenshots on iOS. Signal president Meredith Whittaker warns that OS-level AI agents remain an existential threat to the messenger, and she urges Apple, Google, and Microsoft to document every AI access pathway and let developers opt out.
read more β†’

SageMaker JumpStart Adds Model Leaderboard in Studio

πŸ“Š Amazon SageMaker JumpStart now includes a model leaderboard in SageMaker Studio, letting teams compare foundation models side by side on throughput, latency, and cost. AWS evaluates each model with the same setup on BIG-Bench Hard, GPQA Diamond, and MMLU-Pro. At launch it covers more than 70 LLMs from providers such as Meta, Google, Qwen, IBM, and Arcee. It is available at no extra charge in all AWS commercial Regions where SageMaker Studio is offered.
read more β†’

How Spanner Graph Powers Google Ads Fraud Detection

πŸ›‘οΈ Google's Ads team has moved its graph-based abuse detection platform onto Spanner Graph, replacing slow offline MapReduce batch jobs. The platform now runs multi-hop traversals on live production data with sub-second to single-digit millisecond latency. It spans a graph of more than 500 billion nodes and 1 trillion edges, helping Google uncover coordinated fraud rings in real time. The same architecture is available to enterprises in retail, finance, and security.
read more β†’

Google's Data Simulator: A Wind Tunnel for Data Agents

πŸ§ͺ Google has built an internal data simulator that acts as a wind tunnel for enterprise data agents, generating fully synthetic data estates from a short plain-English description. Each estate includes schemas, rows, documents, chat threads, and a verified answer key. The approach lets teams stress-test agents against realistic organizational messiness without ever touching sensitive production data.
read more β†’

VMware vCenter Zero-Day Exploited for Months Before Patch

πŸ›‘οΈ A suspected China-nexus group, Armory Castle, exploited the critical VMware vCenter flaw CVE-2026-123890 (CVSS 9.8) since early September, over two months before Broadcom's fix. The bug lets unauthenticated attackers obtain an administrator session. Broadcom patched it on November 19 in VMSA-2026-0009 and offers no workarounds. Google and Mandiant urge admins to patch and hunt for signs of compromise.
read more β†’

Google Extends Android Developer Verification Worldwide

πŸ” Starting in January 2027, Google will extend Android developer verification to every country and, over several months, to every install source. Developers who have not verified their identity and registered their apps will lose the normal install experience on certified devices. The rollout begins with seven distributors, including Google Play, Galaxy Store and Xiaomi. An advanced flow will still let experienced users install unregistered apps at their own risk.
read more β†’

ThreatsDay: AI Jailbreaks, Ransomware and Stolen Data Sales

πŸ›‘οΈ This week's ThreatsDay Bulletin shows how attackers rely on the familiar, from fake job offers to lookalike domains, while AI makes threats harder to spot. OpenAI and Google unveiled new cyber-capable models, even as the U.K. AISI found a jailbreak regression in GPT-6.2 Astra that has since been fixed. Meanwhile, ransomware arrests, Kimsuky attacks and an AI-driven stolen data marketplace show that old problems persist.
read more β†’

IDC Note: Cameyo Bridges Legacy Apps to AI-Ready Browsers

πŸŒ‰ A new IDC Market Note by analyst Filippo Vanara examines how Cameyo by Google delivers Windows and legacy applications inside the Chrome browser without replatforming. The report highlights benefits in AI enablement, zero trust security, and lower costs. IDC cites up to a 54% reduction in TCO compared with traditional VDI and DaaS. It positions Cameyo as a bridge technology for phased modernization.
read more β†’

Gemini 4 Neon Brings Frontier AI and Cyber Defense Gains

πŸš€ Google has introduced Gemini 4 Neon, a fast frontier-level model built for software engineering, knowledge work, science, and cyber defense. It ranks #3 on the Artificial Analysis Intelligence Index and posts state-of-the-art results on SEC-bench Pro and WebExploitBench. Priced at $3 per million input tokens with a 1 million-token context window, it is rolling out today across the Gemini API, Antigravity, and the Gemini app.
read more β†’

Judge Approves Google's Purchase of Spirit Airlines Data

βš–οΈ A New York federal bankruptcy judge has approved the sale of former Spirit Airlines employee data to Google for $10 million. The ruling follows Google's agreement to restrict how it uses and deidentifies the information. Judge Sean Lane found the deidentification procedures comply with applicable data protection laws. Privacy experts, however, warn that risks to employees remain.
read more β†’

Genie 4 Arrives in Project Genie With Real-Time Audio

🌍 Google DeepMind has rolled out Genie 4 in Project Genie, its most realistic and interactive world model to date. Eligible Google AI Ultra subscribers can now edit worlds mid-session with new prompts and enjoy native real-time audio that adapts to the scene. The update also adds upgraded Genie 3 features, a Project Genie agent, and a community gallery for sharing and remixing worlds.
read more β†’

Google Cloud Spotlights EMEA Startup Momentum at Slush 2026

πŸš€ At Slush 2026 in Helsinki, Google Cloud is showcasing how EMEA startups are moving from AI experiments to deploying autonomous agents at scale. Companies such as Finova, VIA HealthTech, Chalkie.ai and CDON use Gemini and Google Cloud infrastructure to transform their industries. Google also launched the Sovereign AI Accelerator for Europe and Israel, alongside up to $350,000 in cloud credits for startups.
read more β†’

Sashiko: Agentic AI Code Review Reinforces Linux Kernel

🧡 Google's AI & Infrastructure team built Sashiko, an open-source, agentic code review system powered by Gemini, to ease the review burden on Linux kernel maintainers. It separates bug discovery from verification and runs about 1,200 reviews a day. Data suggests clean reviews carry real signal, with fewer merged commits needing follow-up fixes since its rollout.
read more β†’

Star Blizzard Abuses Cloud Accounts for Espionage Attacks

πŸ•΅οΈ New analysis from Google details how Russian espionage group Star Blizzard (also tracked as Coldriver and Cold Relic) combines social engineering with cloud account compromise. Based on two intrusions at US organizations, the research shows attackers using MFA fatigue, bulk data theft and impersonation of trusted contacts to reach new targets. Google also saw fresh phishing activity as recently as November 2026, so the threat remains active.
read more β†’

Designing durable agentic systems with Antigravity

πŸ› οΈ This article explains how prompt engineering has evolved into systems design and describes building durable agentic loops with Google Antigravity. It defines five tiers of agent workβ€”from simple questions to full systemsβ€”and explains why the harness around models, not the model itself, is the limiting factor. The author outlines verification gates, evaluation loops, durable state, halt conditions, and standardized skills as core pillars to keep long-running agents reliable and debuggable.
read more β†’

Google and Anthropic accelerate vulnerability disclosure timelines

πŸ”’ Google Project Zero is cutting its vulnerability disclosure deadline from 90 to 60 days in February and to 30 days in January 2028, and Anthropic is moving to 30 days for high/critical issues and 45 days for others starting in January. The change responds to AI-driven advances that radically shorten the time attackers need to find and weaponize flaws, including automated pipelines that reverse-engineer patches. Project Zero is also removing its 30-day "soak time," abandoning severity-based thresholds, and will no longer wait for vendor patch verification before publishing details.
read more β†’

Google Warns of Imminent Fully Agentic AI Attacks

πŸ” Google’s Threat Intelligence Group warns that fully autonomous, agentic attacks could emerge within three to six months after observing a shift toward pen-testing frameworks like ARTEX. The GTIG AI Threat Tracker cites increased use of open-source multi-agent frameworks such as CAIRN, CyberStrikeAI and PentAGI, and notes actors are experimenting with custom harnesses. These tools orchestrate sub-agents across multiple kill-chain stages and use LLMs for on-demand payloads, increasing speed and scale while currently trading stealth for velocity.
read more β†’

Google launches Sovereign AI accelerator for Europe

πŸš€ Today Google opens applications for the Google for Startups Accelerator: Sovereign AI, a 15-week equity-free hybrid program available in Europe and Israel. The initiative targets late-seed to post-Series A startups building AI for regulated, sensitive workloads and provides mentorship from Google engineers, regulatory guidance, and partner engagement. Participants gain support deploying on Google Cloud’s sovereign portfolio, including regional data controls, partner-operated clouds, and air-gapped environments, plus access to investors and ecosystem partners.
read more β†’

Google Cloud Announces Expanded Sovereign AI Options

πŸ›‘οΈ Today at the Sovereign AI Summit in Brussels, Google Cloud outlined expanded sovereign cloud capabilities for Europe, addressing data residency, access, and operational control. The company announced Gemini models coming to Google Cloud Dedicated, new cryptographic key control options including Confidential External Key Management (cEKM), and partnerships with local operators and vendors. It also introduced Data Boundary frameworks and Audit Manager support to simplify regulatory audits and compliance for European customers.
read more β†’

Adversaries Scale Attacks with Agentic AI Frameworks

πŸ” Google Threat Intelligence Group (GTIG) and Mandiant report that adversaries are moving from basic prompting to autonomous multi-agent frameworks, automated exploit pipelines, and LLM-focused monetization such as LLMJacking. These developments compress attack timelines, scale intrusions to machine speed, and introduce new supply chain and IO risks. Google outlines defensive measures, model safeguards, and operational mitigations to detect and disrupt agentic threat activity.
read more β†’