AI security tests and software updates shape tech week

It’s been a busy week in tech, with AI and security making big headlines. Google’s AI model Gemini was put to the test in a controlled cybersecurity exercise, where it managed to breach three companies—showing both the power and potential risks of advanced AI. Around the same time, Hacktron researchers uncovered a flaw that let them access OpenAI’s internal code repository, raising concerns about how even leading AI firms protect their own systems. On the software side, OpenAI brought ChatGPT closer to everyday work by adding Microsoft Word support and a way to link multiple accounts through plugins. Apple also launched a beta version of its new Siri AI as part of iOS 27, while Oracle released Java 27 with post-quantum cryptography to prepare for future security threats. Elsewhere, Alibaba shared its Qwen-Image 2.1 model, which can create and edit images with transparent backgrounds, and Google rolled out Android 17 QPR1 with tweaks to Pixel status bar icons. Microsoft rounded out the week with a record-breaking Patch Tuesday, fixing 974 vulnerabilities in one go—the most ever in a single update. Together, these events show how fast AI is evolving, how deeply it’s being woven into our tools, and why security stays a constant race.