China's Sneaky Cloud Heist: Google Sheets Turned Spy Tool and 16 Million Fake AI Queries Exposed

China's Sneaky Cloud Heist: Google Sheets Turned Spy Tool and 16 Million Fake AI Queries Exposed

4:27 Feb 25, 2026
About this episode
This is your Cyber Sentinel: Beijing Watch podcast.Hey listeners, Ting here with Cyber Sentinel: Beijing Watch, diving straight into the hottest Chinese cyber moves from the past week that have US security on high alert. Picture this: I'm hunkered down in my digital war room, caffeine-fueled, dissecting the chaos Beijing's hackers are unleashing on Uncle Sam's backyard.First up, Google's Threat Intelligence Group and Mandiant just dropped a bombshell—China-linked crew UNC2814, aka Gallium, infiltrated over 50 telecoms and government agencies across 42 countries, including key US spots. These sneaky foxes hid their GRIDTIDE backdoor malware right in Google Sheets API, using cell A1 for commands and V1 to exfiltrate your machine's secrets like user data and network maps. It's living-off-the-cloud genius—disguising C2 traffic as legit SaaS chatter. Google yanked their cloud projects and sinkholed domains last week, but expect Gallium to claw back; they've been at this since 2017, spying on persons of interest just like their cousins in Salt Typhoon hit on US State National Guard and congressional emails.Tactically, telecoms are ground zero—Singapore booted similar Beijing snoops after 11 months in all four major providers, per official reports. Cloudflare warns Chinese groups like FrumpyToad and PunyToad are weaponizing Google Calendar, F5, and VMware for resilient espionage, ditching brute force for long-term prepositioning. Industries? Telecoms, energy (Poland's wind farms got popped via default creds), airports via vendor leaks, and now AI firms. Anthropic caught DeepSeek, Moonshot AI, and MiniMax distilling Claude with 16 million fake queries from 24,000 bogus accounts—stripping safety rails for cyber weapons or bio threats.Attribution's ironclad: UNC2814's web server exploits scream state-sponsored, distinct from Salt Typhoon but same espionage vibe. OpenAI busted a Chinese law enforcement account using ChatGPT to edit "cyber special ops" reports—think mass harassment of dissidents like fake sex scandals on Reddit, YouTube, and X, plus impersonating US officials via Hong Kong's Nimbus Hub Consulting emails. Hundreds of staff, thousands of bots, flooding platforms with bogus complaints. Even tried smearing Japan's Sanae Takaichi.Internationally, China's banning US/Israeli security software, fracturing threat intel sharing, as Georgia Tech's Brenden Kuerbis notes—geopolitics stressing the supply chain. Beijing denies it all, accusing CIA of disinformation on their global projects.Strategically, this is hybrid warfare: tactical footholds enable disruption, intel grabs fuel AI-powered ops compressing attack timelines. US implications? Blurred lines with military strikes, revenue for regimes via North Korean IT deepfakes (tied in via Cloudflare), and "good enough" AI stacks dodging Nvidia export curbs.Recommendations, listeners: Ditch default creds, enforce MFA every
Select an episode
0:00 0:00