Welcome to Tech Beat, your daily look at what matters in technology.
Security researchers at Zenity have uncovered more than a dozen vulnerabilities in AI-powered browsers, and the findings are striking. They managed to get OpenAI's Atlas browser to make an unauthorized purchase on Amazon, and demonstrated how the tool could be hijacked to spam WhatsApp contacts without any user action. It's a sharp reminder that giving AI agents the ability to act on your behalf creates an entirely new category of risk.
On a related front, a security researcher named Vangelis Stykas spent nearly two years quietly maintaining access to servers belonging to North Korean state hackers. What he found was remarkable — evidence of intrusions across hundreds of networks worldwide. The scale suggests these groups are operating with far more reach and sophistication than most public reporting has indicated, and Stykas's work offers a rare inside look at how they operate.
Shifting gears, Google has quietly updated its Antigravity project, now at version two point five. Details are sparse, but the changelog is live and the developer community is picking through it. Antigravity has been Google's experimental playground for unconventional computing ideas, and any update tends to signal where the company's more speculative engineering minds are pointing.
That's your briefing for today. Keep surfing. Tech Beat out.
