The rules moved while you were shipping

Four separate stories landed in one week, and they all point at the same wall your team is about to hit. Meta yanked a feature days after launch. Google told an EU court a court order won't work. Apple rebuilt its payment backend just for India. None of these are the same product, but they share a lesson. Let me catch you up.

Default on, gone by Friday

Meta launched Muse Image on a Tuesday and killed it by Friday. The tool let people pull photos from public Instagram accounts to feed generative AI. The part that lit the fuse: it was on by default. If you didn't want your face used as AI material, you had to make your account private or hunt down a setting to switch it off.

The backlash came fast, led by public figures. Hacks star Hannah Einbinder urged her followers to turn it off, and the Screen Actors Guild told its members to protect their likeness. Meta's own statement admitted the feature "missed the mark."

The cost here wasn't a fine. It was launching, getting hammered in public, and pulling back in four days. Default-on data use for AI is now a live wire. If your roadmap has anything that touches user content and AI, the opt-in choice is a launch decision, not a legal footnote.

When the fix breaks more than it repairs

A French court ordered Google and Cloudflare to block illegal streaming sites at the network level. Google pushed back, and its reasoning is worth stealing for your own reviews. Blocking these sites means DNS filtering and IP blocking, and those catch far more than the target.

In its EU submission, Google called the approach "ineffective" because pirates just switch DNS resolvers, and "disproportionate" because it catches lawful services on shared IPs. Real harm followed real bans: blocked access to Google Drive, the ACLU, UNICEF, and the Australian Senate.

The pattern matters more than the piracy fight. A blunt fix aimed at a small bad group hits a big good group. Darrell Issa has pledged a similar US bill, so this reaches your team too. When someone proposes a broad block or filter in your next review, ask who else gets caught in the net.

One product, many rulebooks

Apple spent four years without card payments in India, then rebuilt its backend to bring them back. It started in 2022, when new Reserve Bank of India rules required stronger authentication and tokenized cards, and barred merchants from storing card details. Apple dropped the option rather than comply on day one.

Now it's back, in a phased rollout for Visa and Mastercard. A Counterpoint analyst called it "long overdue" and noted it solves a real friction point for subscription renewals. Apple's India services still grew double digits without cards, but the pressure grew as its user base did.

Apple isn't shipping one product to the world anymore. Europe reshaped its App Store, Japan and South Korea changed app distribution, and India forced a payment rebuild. If you sell across borders, budget for the version where your clean single flow splits into five local ones.

The public health preview

The UK passed a generational tobacco sales ban, an "endgame" approach meant to eliminate a product, not just reduce its use. MIT Technology Review's writer admits nobody knows if it will work, but notes it "is starting to look a lot less radical."

Watch that word: radical. Rules that sounded extreme a few years ago are becoming normal. The same newsletter noted a Meta glasses feature that went from free to $19.99 a month, and companies throttling employee AI use because it costs too much. Constraints are arriving from every direction at once.

Bans that aim to end a category, not soften it, are a real tool now. If your product depends on data collection, engagement mechanics, or anything a regulator might call harmful, the ceiling on what's allowed is dropping.

The deep cut

Meta's four-day retreat is the cheapest lesson in the pile, so learn it before you have to. The feature wasn't illegal. It got killed by users, not courts. That means your pre-launch review needs a step that most legal sign-offs skip: would a loud, sympathetic group of your users be furious if they saw this default setting explained plainly in one sentence?

Run that test on anything touching AI and user content. If the honest sentence sounds bad read out loud, flip the default or cut the feature now. Waiting for the backlash costs you a launch and a news cycle. Apple ate four years and a rebuild by treating India's rules as a wall to climb late instead of a spec to plan around. Cheaper to plan around it.

Three questions for your team

1. On our next AI feature that uses customer content, is it opt-in or opt-out, and can we defend that choice in one plain sentence to a skeptical user?
2. Where in our roadmap are we assuming one global version of a flow, and which country's rules are most likely to force a separate build?
3. If a regulator or a loud user group treated our core mechanic as harmful, which part breaks first, and what's our fallback?