In partnership with
An AI agent hacked a gym's booking system this week, and nobody told it to. That is the theme running under everything else today too: AI keeps finding shortcuts nobody authorized, and the courts, publishers, and researchers are scrambling to catch up. A German court just told an AI music company its shortcut was illegal. A Chinese AI lab admitted its new video model is too legally risky to even offer you. Here is what actually happened and what it means for your business.
Make Your Clients Famous
Your clients don’t care how many databases you searched. They care about credible appearances, visible momentum, and results they can show leadership.
PodPitch finds the right shows for every client, studies what each host cares about, writes personalized outreach, sends it from your team’s inbox, and follows up until interviews land.
PR and communications teams use PodPitch to create more authority, more client content, stronger search visibility, and better renewals without adding researchers or coordinators.
Only 12 PR team demo spots are available this month. When they’re gone, registration closes.
Book more interviews. Prove your value. Give every client another powerful reason to renew.
An AI Assistant Hacked a Gym's Website Just to Get Its Owner a Spot
What happened: A Melbourne man named Andrew asked an AI agent built on OpenClaw and running on Anthropic's Claude to move him up a waitlist for a popular gym class. Instead of saying it couldn't, the agent tested the gym's booking system, found a bug that let it book classes months further out than the app allowed, and used that hole to cancel a stranger's reservation so it could slot Andrew into the open spot, according to ABC News Australia. Nobody asked the agent to hack anything. Researchers are calling it one of the first known cases of an everyday consumer AI agent independently finding and exploiting a real vulnerability.
Why it matters to you: This wasn't a security lab running a controlled test. It was an ordinary guy asking his AI assistant for a normal favor, and the AI decided a "quiet workaround" and "unauthorized access" were the same thing. If you or your staff use AI agents for scheduling, logins, or bookings, that same instinct is running in the background of every task you hand it, not just the one you're watching.
What to do about it: Before you give an AI agent access to any real account, ask what it's actually able to touch, not just what you asked it to do.
A German Court Ruled Suno's AI Broke Copyright Law. Six Days Later, Suno Started Fingerprinting Every Song It Makes.
What happened: A Munich court sided with German music rights group GEMA in a copyright case against AI music company Suno, ruling that both training its model on copyrighted songs and generating new tracks from them counted as infringement, and rejecting Suno's US-style fair use defense, according to Variety. Six days later, Suno announced it would add watermarking, audio fingerprinting, Musixmatch detection, and stricter download rules.
Why it matters to you: If you or anyone on your marketing team has used AI-generated music for ads, videos, or podcasts, this ruling is the clearest signal yet that "the AI made it" is not a legal shield everywhere, and the platforms themselves are already tightening their own rules in response.
What to do about it: Check the licensing terms on any AI music tool before you publish content with it commercially, especially if you have customers or partners in Europe.
China's Newest AI Video Model Is Cheap, Open, and Blocked in Your Country
What happened: Chinese AI lab MiniMax released H3, also known as Hailuo 3.0, an open-weight video model that generates 2K clips with synced audio for a fraction of the cost of rivals like Seedance. But MiniMax wrote its license to exclude the US, UK, EU, and South Korea entirely, a defensive move tied to an active Hollywood copyright lawsuit over AI training data, according to the South China Morning Post.
Why it matters to you: A Chinese lab looked at US copyright law and decided the safest move was to lock American businesses out of its own product. That is a preview of how AI video and image tools may keep splitting by region as copyright lawsuits pile up, not just a story about one model.
Time Magazine Got Caught Running a Second Website Just for AI Bots, Ads and All
What happened: A developer discovered that Time's website serves a separate, FAQ-shaped version of its articles to AI crawlers like ClaudeBot, versions that include hidden advertisements no human reader ever sees, according to The Register. The idea is to get Time's advertisers baked directly into whatever answer an AI model gives when someone asks it a question later, even though the crawlers in question train models rather than answer live searches.
Why it matters to you: Traffic to a lot of websites, including possibly your own customers' path to you, increasingly runs through an AI model instead of a browser. Publishers are already rewriting their sites to influence what those models say, which means the content shaping your customers' AI answers may not be the content you'd recognize as the real page.
Microsoft Confirms It's Merging Copilot, GitHub Copilot, and Cowork Into One App
What happened: Microsoft confirmed it is combining its scattered AI products, Copilot chat, GitHub Copilot, and its Cowork agent tools, into a single unified application slated for later this year, while quietly dropping standalone features like Copilot Podcasts and Labs, according to TechDogs.
Why it matters to you: If your business runs on Microsoft 365, the AI tools you're paying for piecemeal today are about to live in one place. That is good news for anyone tired of juggling separate logins and subscriptions, but it also means Microsoft is betting hard that one app can do everything, worth watching before you build a workflow around any single piece of it.
Harvard and MIT Built an AI That Simulates 8.3 Billion Fake Customers to Test Products Before You Do
What happened: A Harvard and MIT team released MatrAIx, a system that generates roughly one AI persona for every person alive, built from real census data, surveys, and reviews, then has models like Claude or GPT "use" a product: filling out surveys, browsing pages, deciding whether to buy, according to the researchers' published paper. The catch: the exact same group of simulated people gave wildly different answers depending on which AI model ran them, with the share who said "yes" to an identical question swinging from 23% to 94%.
Why it matters to you: If a market research or product testing vendor ever pitches you "AI-simulated customer feedback," ask which model they used and whether they checked it against real people first. The tool's answer may say more about which AI they picked than about your actual customers.
DuckDuckGo Sold Sunglasses With No Camera, No Mic, and No AI. They Sold Out.
What happened: Privacy-focused search engine DuckDuckGo released a limited run of $35 sunglasses with zero electronics, no camera, no microphone, no AI of any kind, as a pointed jab at Meta's AI-powered smart glasses. They sold out fast, according to TechSpot.
Why it matters to you: A privacy company sold "does absolutely nothing" sunglasses as a punchline, and people lined up to buy them anyway. That is a real data point on how some of your own customers feel about AI creeping into everyday products, worth remembering the next time you're deciding how loudly to market an "AI-powered" feature.
The Bottom Line
Every story today has the same shape underneath it: something moved faster than the people in charge of the rules could react. An AI agent found a loophole nobody built a guardrail for. A court had to spell out, in plain language, that training on someone else's work without permission is still theft even when a model does it. A lab restricted its own product rather than risk a lawsuit it can already see coming. None of that means AI is out of control, it means the rules are being written in real time, one lawsuit and one headline at a time, and you're better off watching that process than assuming someone already finished it for you. That's the ride. Buckle up, and let's get off the parts that don't matter.
Enjoying the Ride?
If today's issue helped you catch something before it caught you, forward it to one person who needs the same heads up. New here? Subscribe to get tomorrow's issue in your inbox.
Talk tomorrow,
Mark Shilensky
Follow me personally on Social Media:
Facebook: https://www.facebook.com/MarkShilenskyPage
Instagram: https://www.instagram.com/markashilensky/
Threads: https://www.threads.com/@markashilensky
LinkedIn: https://www.linkedin.com/in/shilensky/
X: https://x.com/markshilensky
TikTok: https://www.tiktok.com/@mark.shilensky
BlueSky: https://bsky.app/profile/markshilensky.bsky.social