Google has launched Gemini 2.5 Deep Think, its most advanced AI model to date, now available to Google AI Ultra subscribers through the Gemini app. The model employs parallel thinking and extended inference time to address complex reasoning tasks, including mathematical conjectures, advanced coding, and design challenges. Performance benchmarks include a 34.8% Human-Level Evaluation (HLE) without tools, 86.6% on Live Code Bench, and 99.2% on the AIME 2025 math competition. Additionally, Google introduced Jules, an AI coding agent powered by Gemini 2.5 Pro, which has recently exited beta and is designed to be fast, visual, and accessible beyond developers. In cybersecurity developments, Google disclosed that its AI-based bug hunter, powered by Gemini, identified 20 security vulnerabilities. The company also confirmed a data breach in June involving unauthorized access to contact information of small business clients via its Salesforce CRM instance. The breach was linked to the threat group UNC6240, associated with the ShinyHunters hacking collective, which used voice phishing techniques to gain access. Google has taken steps to mitigate the impact. Meanwhile, SonicWall is investigating a potential zero-day vulnerability in its Gen 7 firewalls exploited by Akira ransomware in over 20 confirmed attacks. Trend Micro reported active exploitation of critical zero-day flaws in its Apex One security software, prompting patches to address these vulnerabilities.