ThePlan™ has been (and still is for many!) to do a lot of evaluations and maaaaybe force transparency onto AI corporations. ThePlan™ was designed by people who tried to look as respectable as possible, instead of properly engaging with the reality of the risks from AI. By design, ThePlan™ could never address the recent misalignment episodes, superintelligent systems, open source, or any of the near-term risks. Today's near-term risks are yesterday's long-term ones. Back then, they sounded futuristic and not respectable. Thus, ThePlan™ largely ignored them, in order to look more respectable. Now, it is obvious that ThePlan™ was not up to par. It helped a few organisations become more respectable, but it absolutely did not help our institutions with addressing any of the actual risks. So, enough with ThePlan™! I recommend instead The Proper Plan. The Proper Plan is to work with our governments to negotiate international agreements that globally ensure that we: 1) Prohibit the development of superintelligent AI systems, which we lead to actual catastrophes in the case of a misalignment episode akin to the ones we witnessed. For inspiration on how to make it concrete: look at the Chemical and Biological Weapon Conventions. 2) Constrain and regulate frontier AI research. For inspiration on how to make it concrete: look at the protocols around gain-of-function research. 3) Criminalise the publication, the sharing and the hosting of critical AI artefacts. These include the open-weights of agentic models, SOTA code, papers on post-tuning methods, and so on. For inspiration on how to make it concrete: look at the enforcement stack that already exists to suppress digital information, like the obligations and liability on software and hosting platforms around scanning for and taking down CSAM. Without #1, a misalignment episode becomes an irreversible global catastrophe, up to humanity's extinction. Without #2 and #3, we cannot realistically enforce 1. Because ThePlan™ was to let the situation worsen, we are now in a bad spot. As a result, The Proper Plan now costs more than it would have 3 years ago. On top of this, we have little time to prepare. Thus, enacting The Proper Plan will necessarily involve some accidental overcorrections. This is sadly the state of the world we live in. We can clearly make it. But it's a pain: we will need to put a fair amount of effort and iteration.