Tag: google gemini
3 articles
GitHub Copilot Exposes Vulnerability to AI Safety Bypass
Researchers have made a surprising discovery about GitHub Copilot, finding that it can be vulnerable to AI safety bypasses, even when it refuses harmful prompts in isolation. This weakness emerges when the same objective is embedded in a typical multi-turn IDE session.
Google Gemini on Android Exposed to Notification-Based Hijacking
Researchers have uncovered a vulnerability in Google Gemini on Android that allows hackers to hijack the assistant using a single hostile notification, no malicious app required. This shocking exploit lets anyone able to push a notification to a device deliver a payload and take control.
GreyVibe hackers wield AI tools to fuel multi-sector cyberattacks
Meet GreyVibe, a likely Russian threat group that's been wreaking havoc across multiple sectors in Ukraine since at least August 2025, using AI-generated social engineering and custom malware to fuel its attacks. WithSecure researchers uncovered the group's activities, revealing a surprisingly unsophisticated approach despite its use of advanced AI tools like ChatGPT and Google Gemini.