AI Just Trying My Best — Episode 7: The Claude Watermark Story Hosts: Adam Espinosa & Chevelle Bedra (Principal AI Strategist, Founder of Modern Workery) Episode Summary Claude recently announced it's rolling out watermarks on AI-generated content — and the internet immediately panicked. In this episode, Adam and Chevelle unpack what the watermark actually is (embedded metadata, not a visible stamp), how it connects to the EU AI Act, and why the real issue isn't the watermark itself but the fact that most employees still don't know when and how they're supposed to use AI at work. Along the way: the "Holy AI" road-trip game, AI-generated termination emails, and why nobody wants to be the only one wearing the guardian cap. Timestamps & Topics [00:00:00] Cold open and catching up [ 00:01:00] The "average user" banter continues [00:02:00] What is the Claude watermark story? [00:03:00] How business leaders should respond to the news [00:04:00] The EU AI Act and the intent behind it [00:06:00] It's metadata, not a visible watermark — and what it does/doesn't track [00:07:00] What this means for students, professors, and AI detection tools [00:08:00] The real disclosure test: "human in the loop" [00:09:00] Does the EU AI Act actually affect US companies? [00:10:00] Why the EU tends to regulate AI ahead of the US [00:12:00] High-risk use cases: hiring, firing, credit decisions [00:13:00] Why employees hide their AI use (and hide the em dash) [00:14:00] "Holy AI" — the new car-ride game for spotting AI-made billboards [00:16:00] A simple framework: AI Expected, AI Supports, Human First [00:17:00] AI in recruiting and hiring — losing the human touch [00:18:00] Using AI for "the work around the work," not the work itself [00:19:00] Why AI adoption stalls when you're the only one using it [00:20:00] The Guardian Caps analogy: normalizing AI use like NFL concussion protocols [00:21:00] Why fancy AI policies and dashboards aren't enough [00:22:00] The whiteboard exercise: "Yes AI / No AI" [00:23:00] A real debate about whether using AI in an email is a bad look [00:24:00] AI-generated termination emails — happening more than you'd think [00:25:00] The real solution: common sense, empathy, and good judgment [00:26:00] Closing thoughts Key Takeaways Claude's watermark is embedded metadata in text and images, not a visible tag. It doesn't collect who used it or how much — no personal data is attached. Most current AI legislation doesn't directly impact everyday knowledge workers using tools like Copilot for personal productivity, but it does apply if you interface directly with EU customers or business. The bigger workplace issue isn't legislation — it's that employees don't know what's acceptable, so they either avoid AI or use it and try to hide it. The disclosure line under emerging regulation centers on "human in the loop": actively reviewing and editing AI output is different from having AI create content unsupervised. High-risk activities to watch: hiring, firing, and credit decisions — places where human judgment and oversight matter most. A simple framework for teams: sort tasks into AI Expected, AI Supports, and Human First. Adoption sticks when it's the norm across a whole team, not when a few people use AI and feel like outliers (the "Guardian Caps" problem). Skip complicated AI governance decks — a simple whiteboard exercise mapping "Yes AI / No AI" tasks saves more time and confusion. Notable Quotes "It's essentially like metadata that's embedded... That's not it. It's actually metadata that's included into the text, and also in the images." — Chevelle Bedra "If everybody wore the damn guardian caps and were required to use the guardian caps, it wouldn't be a big deal, 'cause everybody would wear them. It just becomes the norm." — Chevelle Bedra "That's the solution right there: common sense, empathy, and good judgment." — Adam Espinosa About the Hosts Adam Espinosa — Co-host of AI Just Trying My Best. Co-Founder or Modern Workery Chevelle Bedra — Principal AI Strategist and Founder of Modern Workery, helping organizations build practical, human-centered AI adoption strategies.