Poetic Prompts Hack AI Safety; Mistral's Open Models; CTV Ad Surge; OpenAI's Garlic Advances

In this episode of AI in Marketing, we delve into the latest breakthroughs and challenges in the AI landscape. Researchers from Sapienza University of Rome, the Sant’Anna School of Advanced Studies, and Dexai have uncovered a method to bypass large language models' safety measures using poetic prompts, significantly increasing the likelihood of generating unsafe outputs. Meanwhile, French startup Mistral introduces its Mistral 3 family of open-weight models, aiming to enhance AI accessibility and enterprise service with customizable and efficient solutions. We also explore how companies are leveraging agentic AI to revolutionize supply chain management, transforming it into a continuous, autonomous process. In the advertising realm, AI-generated ads are gaining traction on connected TV screens, promising rapid ad creation and targeting small- to medium-sized businesses. Despite economic concerns, U.S. consumers set a record with $44.2 billion spent during Cyber Week 2025, highlighting the enduring appeal of holiday deals. OpenAI is developing a new language model, codenamed Garlic, to compete with Google's advancements, marking a strategic push in the AI race. In corporate shifts, Apple's AI leadership sees a change as John Giannandrea steps down, with the company aiming to enhance its AI capabilities under new leadership. Finally, Senator Bernie Sanders calls for congressional action to address AI's rapid development and its societal implications, emphasizing the need for democratic oversight and proactive measures. Stan Berteloot is a French journalist, marketing strategist, and AI expert based in Princeton, New Jersey. A University of Maryland journalism graduate, he navigates the intersection of storytelling, technology, and culture. As Head of Strategy and CTO at Nytro Marketing, he pioneered AI-driven content creation, launching the AI in Marketing podcast. He also hosts Back in America, a podcast exploring American identity through in-depth conversations on race, misinformation, and AI ethics. Start your VoiceStream free trial today! Stan Berteloot is a French journalist, marketing strategist, and AI expert based in Princeton, New Jersey. A University of Maryland journalism graduate, he navigates the intersection of storytelling, technology, and culture. As Head of Strategy and CTO at Nytro Marketing, he pioneered AI-driven content creation, launching the AI in Marketing podcast. He also hosts Back in America, a podcast exploring American identity through in-depth conversations on race, misinformation, and AI ethics. Start your VoiceStream free trial today! 📚 Sources: https://www.darkreading.com/threat-intelligence/researchers-use-poetry-to-jailbreak-ai-models https://techcrunch.com/2025/12/02/mistral-closes-in-on-big-ai-rivals-with-mistral-3-open-weight-frontier-and-small-models/ https://www.pymnts.com/artificial-intelligence-2/2025/agentic-ai-turns-supply-planning-into-a-continuous-autonomous-system/ https://www.marketingbrew.com/stories/2025/11/25/ready-or-not-ai-generated-ads-have-come-to-ctv https://neilpatel.com/blog/boost-app-success-seasonal-user-acquisition-campaigns/ https://au.investing.com/news/stock-market-news/vistra-corp-upgraded-to-investment-grade-by-sp-on-improved-risk-profile-93CH-4152104 https://www.fastcompany.com/91452240/apple-ai-chief-john-giannandrea-dismissed-google-ai https://www.theguardian.com/commentisfree/2025/dec/02/artificial-intelligence-threats-congress