Silly season is almost upon us, and the team responds with an episode brimming with AI news and insights.

What happens when one of the world's leading cryptography organisations loses the key to its own election? Or when an AI model produces more security vulnerabilities because you mentioned Tibet? This week's Priviso Live tackles the fascinating intersection of artificial intelligence, security failures, and the accelerating arms race between cyber attackers and defenders.

Hosts Lyn, Stephen, and Kayla dive into CrowdStrike's alarming discovery about DeepSeek-R1, a Chinese AI coding model that generates up to 50% more security flaws when processing politically sensitive topics. The implications extend far beyond China's borders, raising critical questions about which AI tools South African businesses should trust—especially with municipal elections on the horizon.

From there, the show explores an ironic twist: the International Association of Cryptologic Research had to cancel its leadership election after losing the decryption key. Even the world's top security experts aren't immune to basic key management failures—a sobering reminder as electronic voting systems gain traction.

The episode also covers OpenAI's split with analytics provider Mixpanel following a data breach, Google's launch of the powerful Gemini 3 model, ChatGPT's new group collaboration features, and crucial AI security best practices for 2026. With Google predicting that AI will become standard equipment for both attackers and defenders, understanding these emerging threats isn't optional—it's essential.

Whether you're a developer using AI coding assistants, a business leader evaluating new tools, or simply concerned about deepfakes in the upcoming election cycle, this episode delivers the insights you need to navigate our rapidly evolving digital landscape.

*Subscribe to Priviso Live on Apple, Spotify, or YouTube—because staying informed is your first line of defense.*