In December I spoke on a panel on AI regulation, offering an in-house counsel’s perspective on the ever-increasing and ever-changing world of laws and rules applicable to AI. There were lots of useful nuggets from the panel, and I spent time talking about 1-way vs 2-way doors. One-way doors are decisions that are hard to reverse; you should be very deliberate when making a decision about a 1-way door. Two-way doors are decisions that can be reversed; you should be able to make these decisions more quickly, with less people, information, and process, because you can always come back through the door and make a different decision. This concept is closely associated with Jeff Bezos and Amazon though not exclusive to them.

When it comes to an area where there is great uncertainty, like AI regulations, it’s tempting to treat everything as a 1-way door. Information is imperfect, the landscape is constantly shifting, and a mistake feels consequential. But the reality is that almost every decision in this area (and most every decision really) is a 2-way door. Regulators are figuring out the technology and the rules too, and there’s often opportunity for discussion and remediation even if you do step wrong so long as you acted reasonably and with good intent (and I’d say that doing right by your customers is always reasonable and well-intentioned).

Since the panel, as I’ve considered some of the decisions facing me, I’ve been reminding myself that I’m usually standing in front of a 2-way door. It’s allowed for more decisive decisionmaking and increased comfort with making a mistake, knowing that I can cure it later if necessary. I think this is a useful rubric that anyone can use for any decision, but I’ve found it particularly helpful in thinking through decisions and avoiding paralysis where we have less information than we’d like.