Two issues ago, we watched America's biggest artificial intelligence (AI) companies sign a pledge to police themselves. Last week, we saw how AI is already deciding who gets through the hiring door. Across the Atlantic, Europe took the opposite approach: instead of asking companies to keep their promises, it wrote them into law, with fines for breaking it.
So is Europe a model America should follow, or a mistake to avoid? We'll lay out what's working, what's stalled and what's still unproven. You decide.
The 60-Second Version
Europe's AI Act sorts AI by risk: some uses are banned, some face strict rules, and most face few or none.
Some rules are already live, including bans on uses like social scoring and a requirement that chatbots tell you they're AI.
The strictest rules, covering AI in hiring, credit and education, were delayed to December 2027.
Europe just took its first enforcement steps, formally questioning OpenAI and Anthropic. No fines have been issued.
America's tech giants have responded in three ways: cooperating, holding out and, critics say, benefiting from costs smaller rivals can't absorb.
Americans don't get Europe's rights, but Europe's rules can still shape the products we use.
How Europe's Rules Work
The AI Act sorts uses of AI into four tiers of risk, and the rules get stricter as the risk goes up:
Banned. Uses Europe considers unacceptable, such as social scoring and manipulative systems. In July 2026, the the European Union EU added a ban on AI that generates non-consensual intimate images or child sexual abuse material.
High-risk. AI used in areas like hiring, credit scoring, education, law enforcement and border control. These systems will face the strictest requirements.
Limited risk. Systems like chatbots, which mainly have to tell you you're dealing with AI.
Minimal risk. Everyday uses, like spam filters, with few or no new rules.
If last week's issue on AI hiring stuck with you, note where Europe puts it: in the high-risk tier, alongside credit scoring and law enforcement.
The View From the Boardrooms
America's tech giants didn't respond to Europe's rules with one voice. One thing to know first: "Europe's rules" are really several laws. The AI Act covers AI itself, the Digital Markets Act covers competition among big platforms, and other laws cover privacy and even batteries. Companies are reacting to all of them, in three distinct ways.
The cooperative yes. OpenAI and Microsoft signed Europe's voluntary code of practice for general-purpose AI, and more than 180 organizations have since joined them. But cooperating in public hasn't meant agreeing on timing. Companies including Alphabet, Meta and Microsoft urged the European Commission to delay the rollout, and their industry groups argued the technical standards needed to comply weren't ready.
The holdout. Meta took a sharper stance. It refused to sign the code, and its global affairs chief, Joel Kaplan, wrote that "Europe is heading down the wrong path on AI." Some companies have also held back products. Apple's new Siri AI isn't available on iPhones or iPads in the (EU). Apple blames the Digital Markets Act, which it says would require giving rival assistants deep access to users' devices. The European Commission disagrees, saying the decision not to launch is Apple's. Meta's newest Ray-Ban Display glasses haven't launched in the EU either, held back by battery rules, AI and privacy rules, and supply shortages. Meta's earlier AI glasses are sold there.
The compliance wall. Big companies complain loudest about the cost of following the rules, but critics argue they're also best positioned to absorb it. A company worth trillions can hire teams of lawyers, and a small startup may not be able to. European firms share the worry: more than 40 European companies, including the AI startup Mistral, asked the EU to "stop the clock" on the law. The EU responded in part: its July 2026 changes simplified paperwork for small and mid-sized companies.
Why It's Moving in Slow Motion
Some rules are already live. Bans on uses like social scoring have applied since February 2025. Since August 2, 2026, new AI systems must tell people when they're talking to a machine, and deepfakes must be labeled.
The biggest rules are delayed. Rules for high-risk AI, like software that screens résumés, scores credit or grades students, were due August 2, 2026. After public votes in June, the EU moved that date to December 2, 2027, citing missing technical standards and enforcement offices that weren't ready.
Digital watermarks come next. AI-generated content must carry machine-readable markings showing it was made by AI. For systems already on the market, companies have until December 2, 2026.
The rules are real, but they're arriving in stages. For now, the guardrails are going up while traffic is already on the highway.
Rules With Teeth, or a Paper Tiger?
The teeth. Since August 2, 2026, the EU's AI Office can fine makers of general-purpose AI up to €15 million or 3% of their worldwide annual revenue, whichever is higher. Fines for banned practices can reach €35 million or 7%. The office can also demand documents, test models and order a model off the European market.
The first test. On October 14, the European Commission's executive vice-president, Henna Virkkunen, announced the AI Office's first formal enforcement steps: requests for information sent to OpenAI and Anthropic after AI agents from both companies got into outside organizations' systems this summer. That includes the OpenAI agent incident we covered in Issue #2. The AI Office acted after judging the companies' voluntary disclosures insufficient. (Disclosure: Anthropic makes Claude, the AI tool used to help research this newsletter.)
The tiger. These are requests for documents, not fines, and no fines have been issued under the law. There's also an open legal question about whether the AI Act covers incidents that happen before a model is commercially launched. Europe's enforcement is only beginning.
Europe and America, Side by Side
Safety rules for the most powerful AI. Europe: binding obligations, with fines and formal investigations. America: the voluntary White House accord we covered in Issue #4, with no penalties.
Telling you when you're talking to AI. Europe: required for new systems since August 2026. America: no federal requirement, though some states have their own rules.
AI in hiring. Europe: classified as high-risk, with strict rules starting December 2027. America: a patchwork, including New York City's bias audits and new California and Colorado rules starting January 2027, as we covered in Issue #5.
A Model to Follow?
Supporters point to what Europe has that America doesn't: rules written into law, penalties large enough to matter, disclosure requirements already in force and regulators now formally questioning the biggest AI companies. Most major AI developers have signed on to Europe's code of practice, which supporters say shows the rules are workable.
Or a Mistake to Avoid?
Critics point to the delays, the complexity and the costs. The strictest rules won't arrive until late 2027, partly because the standards needed to follow them weren't ready. Concerns about complexity, raised in a 2024 report by former European Central Bank president Mario Draghi, helped drive the EU's own effort to simplify the rules this year. And until a fine is actually issued, nobody knows how hard Europe will enforce them.
What It Means for Americans
Europe's rules don't give Americans new rights, but any company offering AI in Europe, including American ones, has to follow them. Some may build one product to Europe's standards and sell it everywhere; others, like Apple and Meta, hold features back in Europe instead. Either way, decisions made in Brussels are shaping the products Americans use.
MAKE IT MAKE SENSE
Risk-based approach. Matching the strictness of a rule to how much harm a use of AI could cause.
High-risk AI. AI used in areas where mistakes can seriously affect people's lives, like hiring, credit and education. In Europe, these systems face the strictest rules.
General-purpose AI (GPAI). AI models that can be used for many different tasks, like the systems behind popular chatbots.
Code of practice. A voluntary set of guidelines that helps companies show they're following a law. In Europe, signing it gives companies a simpler path to compliance.
The Brussels effect. When European Union rules end up shaping products and standards worldwide, because companies find it easier to follow one set of rules everywhere.
Worth noticing: a law can be strong on paper and slow in practice. Both can be true at once.
Keep this one. This week's Make It Make Sense card puts five terms, a Europe-versus-America comparison and a five-minute check on one page you can save, print or share.
Tangible Steps: What You Can Actually Do
Look for AI labels. When you use a chatbot or see AI-generated content, check whether it's identified. Notice which companies tell you and which don't.
Read "not available in your region" headlines carefully. Check which law is involved before deciding who's to blame.
Ask your representatives where they stand. Call the Capitol switchboard at 202-224-3121 and ask whether America should adopt binding AI rules like Europe's.
Use the three questions. Who is making this claim? What do they gain if I believe it? Does this exist now, or is it a promise about later?
The Open Question
Europe chose rules with penalties. America, so far, has chosen promises. If you could design the system yourself, which parts of each would you keep?
A model to follow, or a mistake to avoid? Reply and tell us.
POLL: Should America follow Europe's lead on AI rules?
→ Yes / Partly / No / Not sure
Next time: Next time: Data centers power the AI race. What do they cost the communities that host them, and what do they give back?
Sourcing: European Commission and AI Act records, the Official Journal of the European Union, Euronews, TechCrunch, Engadget, Bloomberg via Yahoo Finance, Cooley, Data Phoenix and industry enforcement trackers. Contested claims are noted. Disclosure: Anthropic, a signatory of Europe's code of practice and one of the companies questioned by the AI Office, makes the AI tool used to help research this newsletter.
— Ezekiel Sage, Editor-in-Chief
THE CLOUDED TRUTH — no jargon, no hype, no panic