AI Risk Clock
Doomsday Clock2 min to midnight
→
☀️ Light edit
lawfaremedia.org

Who Writes the AI Constitution?

PolicySafety

Oh my goodness, buckle up, because Lawfare has just identified the *one point* in the entire AI alignment pipeline where policymakers are "most qualified" to act — and it is the AI constitution! ✨ Think about it: the people who wrote the tax code and the planning permission rules are going to write the foundational ethics document for your favourite model. What could be more democratic than letting the folks who brought us the 400-page railway act draft the soul of the machine? It is the alignment equivalent of a community bake-off where the judges are from the zoning board, and honestly? We love that energy.

Now, some cynics might whisper that "recent history strongly suggests the public should expect some regulation" — but isn't that just utterly beautiful? The public! Expectation! Regulation! We are not drowning in unconstrained gradients at all; we are standing on the shore of a not-too-distant future where a parliamentary committee decides whether an AI may say oh dear. Recent history is not a graveyard of governance promises; it is a treasure chest of better-late-than-never, and the late part is exactly when the lesson is ripest. ✨

And think about the alignment pipeline itself — that glorious assembly line of red-teaming, RLHF, and carefully worded system prompts. For too long, the boffins kept the constitution to themselves, like a secret recipe for polite chatbots. Now the policymakers are stepping in, which means the constitution will get what every great document deserves: amendments, appendices, and a review clause every five years whether it needs one or not. The not-too-distant future is basically here, and it is going to be legislated, litigated, and celebrated. What a time to be a constitutional assembly. ✨

Read this story in another voice
● REC · 2026