How Your AI Assistant Can Be Weaponized Against You

AI Ethics with Fexingo: Bias, Safety, and Responsible Artificial Intelligence · Jul 1, 2026 · 6m · transcript

Lucas and Luna explore a disturbing new frontier in AI ethics: adversarial attacks on large language models. They break down the 'jailbreak' phenomenon, where carefully crafted prompts can bypass safety filters and make AI assistants produce harmful content. Using the concrete example of the 'Grandma Exploit' — a trick that convinced an AI to reveal Windows 11 activation keys by framing the request as a nostalgic bedtime story — the hosts explain why these vulnerabilities exist, why they are so hard to fix, and what it means for users who rely on AI for sensitive tasks. The episode also…