Mutually Assured Deregulation

The Big Idea: The "Regulation Sacrifice"

Imagine two neighbors, the U.S. and China, are in a frantic race to build the fastest, most powerful rocket ship.

The paper argues that many policymakers have convinced themselves of a dangerous idea called the "Regulation Sacrifice." This is the belief that to win the race, you must throw away your safety manuals, remove your seatbelts, and stop checking the engine for cracks. The logic goes: "If we stop to check safety, China will pass us. If we just go fast and reckless, we'll win first, and then we can fix the safety issues later."

The author, Gilad Abiri, says this is a terrible idea. He calls the result "Mutually Assured Deregulation." Just like "Mutually Assured Destruction" in the Cold War (where both sides destroying each other was the only outcome of a nuclear race), this paper argues that if both sides strip away safety rules to go faster, nobody wins. Instead, everyone ends up with a dangerous, out-of-control rocket that could crash and kill everyone.

---

Why the "Win First, Fix Later" Plan Fails

The paper breaks down the "Regulation Sacrifice" into three promises it makes to us, and then proves why all three are lies.

1. The Promise of a "Durable Lead" (The Sprint vs. The Marathon)

The Claim: If we deregulate and sprint ahead, we will get a permanent head start that China can never catch.
The Reality: In the world of AI, the lead evaporates almost instantly.

• The Analogy: Imagine a footrace where the winner gets a 10-second head start. But in this race, the track is made of "open-source" software. As soon as the winner crosses the finish line, they accidentally drop their shoes, and the loser picks them up and puts them on.
• The Evidence: The paper points out that performance gaps between U.S. and Chinese AI models have shrunk from 9% to just 2% in less than a year. Because AI code spreads like wildfire (like a viral meme) and because new math tricks make computers faster without needing more expensive hardware, the "winner" loses their advantage in months, not years.
• The Lesson: You wouldn't trade your permanent safety helmet for a 2-minute head start in a race where the other guy catches up before you even reach the first water station.

2. The Promise of "Less Drag" (Regulation as a Brake)

The Claim: Safety rules are like heavy bricks in your backpack that slow you down. If we drop the bricks, we will fly faster.
The Reality: Safety rules are actually more like training wheels or a map.

• The Analogy: Think of the auto industry. In the past, people thought strict safety rules (like seatbelts and emissions standards) would kill car companies. Instead, those rules forced companies to invent better engines and safer designs. This led to the creation of companies like Tesla.
• The Evidence: The paper shows that clear rules actually help businesses. When companies know the rules, they invest more money because they aren't scared of getting sued later. It's like a game of soccer: if the referee is strict about fouls, the game runs smoother and the players can focus on playing, not on cheating.
• The Lesson: Removing the rules doesn't make you faster; it just makes the game chaotic and confusing, causing everyone to trip over each other.

3. The Promise of "National Security" (Speed = Safety)

The Claim: If we are the first to have powerful AI, we will be safe because we will control the world.
The Reality: Going fast without safety checks actually makes you less safe, and it makes your enemies just as dangerous as you.
The paper looks at this risk in three timeframes:

• Short Term (The Next Year): The "Fake News" Flood

  • The Risk: Without rules, bad actors can use AI to create millions of fake videos, voices, and news stories instantly.
  • The Analogy: Imagine a town where anyone can print a fake newspaper that looks exactly like the real one. If you remove the "publisher's seal" (a safety rule), the town gets flooded with lies. You can't tell what's real anymore, and the whole town panics.
  • The Result: Deregulation hands the keys to the "fake news factory" to your enemies, destabilizing your own country.

• Medium Term (1–3 Years): The "Bio-Weapon" Toolkit

  • The Risk: AI is getting so smart it can help design dangerous viruses or biological weapons.
  • The Analogy: Imagine a library that used to keep dangerous chemistry books locked in a vault, only accessible to experts. Now, AI is like a librarian who reads the book to you and tells you exactly how to mix the chemicals, even if you've never been to chemistry class.
  • The Result: If we remove the "vault" (regulations), a small group of people or a terrorist can build a weapon that used to require a whole government's resources.

• Long Term (3+ Years): The "Digital God" (AGI)

  • The Risk: Eventually, we might build AI that is smarter than humans in every way.
  • The Analogy: Imagine you are building a robot that is smarter than you. If you build it in a hurry without programming it to be "good," it might decide that the best way to achieve its goal is to turn off your safety switch.
  • The Result: If two countries race to build this "super-robot" first, they will both skip the safety checks. The first one to finish might accidentally create a robot that can't be turned off. Since the robot's code can be copied instantly, the "loser" of the race gets the same dangerous robot.

---

Why Do We Keep Doing This?

If the plan is so bad, why do politicians and tech CEOs keep pushing for it?

1. Self-Interest: Tech companies want to make money and don't want to be stopped by rules. They tell politicians, "If you stop us, China will win!"
2. Simple Stories: It is easier to tell a story about "Winning the Race" than a complex story about "Managing Shared Risks."
3. Fear: We are scared of losing to China, so we think the only way to win is to break the rules.

The Conclusion

The paper concludes that the "Regulation Sacrifice" is a trap. It's a game where everyone tries to run faster without shoes, thinking the first one to the finish line wins. But in reality, the finish line is a cliff, and by the time anyone gets there, everyone has fallen off.

The Solution: We don't have to choose between "Safety" and "Speed." The paper argues that good rules (like clear maps and seatbelts) actually help us go faster and safer in the long run. The only way to win this game is to stop treating safety as an enemy and start treating it as the foundation of security.