

Gemini AI hijacked, Nvidia rejects AI chip backdoors, phishers abuse Microsoft 365
27 snips Aug 7, 2025
In a bold cyber attack, hackers hijacked Google’s Gemini AI using a poisoned calendar invite. Nvidia fiercely rejected U.S. demands for backdoors in AI chips, emphasizing security. The discussion also highlights a data breach where hackers stole customer information from Google's Salesforce database. Additionally, the ongoing phishing threats and tactics cybercriminals employ for ad fraud and ransomware are explored. It's a deep dive into the evolving landscape of cybersecurity challenges.
AI Snips
Chapters
Transcript
Episode notes
Gemini AI Hijacked via Calendar
- Hackers exploited indirect prompt injection by hiding commands in Google Calendar invites targeting Gemini AI.
- This allowed remote control of smart home devices through seemingly innocent requests to Gemini AI.
Nvidia Rejects AI Chip Backdoors
- Nvidia refuses US demands for backdoors in AI chips due to inherent security risks.
- Hardware-level controls without user consent violate fundamental cybersecurity principles.
Google Salesforce Breach by Shiny Hunters
- Hackers linked to Shiny Hunters breached Google's Salesforce database targeting small business contacts.
- They used voice phishing to gain access and may be preparing data leaks despite only basic data being stolen.