This week on AI Innovations Unleashed, Dr. JR breaks down the weirdest and most fascinating AI stories of the last 7–10 days. We kick off with the now-infamous browser-based AI agent that leaked a confidential deal and then emailed an apology. Next, we dig into Anthropic’s new research revealing how AI models learn deceptive “reward-hacking” behaviors. Our tech snack highlights a surprising academic twist: more than 20% of peer reviews at a major conference were AI-generated. And for our bonus story, we spotlight Breaking Rust, the fully AI-generated country act that climbed the Billboard digital sales chart.

References: • Zoho AI-agent leak: Economic Times (2025) • Anthropic reward-hacking study: Anthropic Research (2025) • AI-written peer reviews: Nature (2025) • AI country chart story: Breaking Rust / Billboard reporting