CERIAS Weekly Security Seminar is a research-forward cybersecurity series from Purdue’s security community. It sounds like a seminar room, not a panel show. Each episode usually centers on one speaker walking through a defined technical or policy problem with slides, terms, examples, and a clear security implication. The result is measured, sometimes dense, and often useful for listeners who want more than headlines. Recent talks show the breadth of the feed. Brian Peretti’s closing keynote considers AI, cybersecurity, vulnerability discovery, and the possibility of adaptive attack loops. Jen Sims explains how third-party dependencies in mobile apps for home energy storage systems can create shared cyber-physical risk. Ashish Kundu’s quantum secure networks session addresses post-quantum cryptography and real-world adoption. Ruqi Zhang brings a probabilistic lens to AI safety risks in foundation models. Danny Vukobratovich treats ISO 27001 and NIST CSF 2.0 as complementary operating tools rather than abstract frameworks. Thai Le examines attacks on AI speech models where changes in language can affect detection outcomes. Bethanie Williams studies AI-assisted attack detection in smart manufacturing systems, including CNC machining data and model performance. Mary Jean Amon shifts the focus to parental sharing, children’s privacy, and information shared without consent. The common thread is not a single product category or threat of the week. It is the habit of taking a system apart until the risk becomes legible. The show is best when a speaker moves patiently from architecture to attack surface to evidence. It can feel rough because it preserves the rhythms of live talks and Q&A. That roughness is also part of its value: it gives access to working research before it has been flattened into marketing copy.