Outsider, a Chinese phishing network, abused Gemini to mass-produce fraudulent SMS messages and websites, caused $1.9 billion in damages, and was shut down through U.S. law enforcement action.
Gemma 4 family with three variants (31B dense, 26B-A4B MoE, E2B compact) is available as a fully managed service on Amazon Bedrock, with native reasoning, function calling, and multimodal support.
Attackers can inject malicious commands into messenger messages through fake context alignment, which Gemini processes undetected and uses to control authorized devices or misuse data.
Google releases Gemma 4 12B as an Apache-2.0-licensed multimodal model with unified architecture that runs locally on laptops with 16 GB VRAM and combines text, image, audio, and reasoning.
A manipulated notification via WhatsApp, Slack, SMS, Signal, Instagram, or Messenger could hijack Google Gemini on Android devices and force it to execute arbitrary actions without requiring a malicious app to be installed on the phone.
Prompt injection vulnerability in Google Gemini Voice Assistant enables hidden malicious commands through manipulated notifications, potentially leading to social engineering and data misuse.
Microsoft has introduced MAI-Thinking-1, its first reasoning model with fine-tuning capability for enterprise, specifically designed for domain-specific customizations.
Google is shifting the focus of its AI platforms from assistive functions to independently operating systems, making mobile and web development a priority in the process.
Open models are closing the gap to the frontier, but different benchmarking methods and evaluation frameworks make reliable performance comparisons between open and closed systems difficult.
