Four filters for every tool

RAND’s 2026 strategy is not a pile of bans. It is four tests. A tool that fails a test is left on the shelf, even if it looks tough.

Prevention over mitigation. Act before the first living cell, because cleanup is not a backup.

Universal restraint over partial restraint. One success anywhere is a problem everywhere. “Most countries promised” is not enough. The reachable version, for now, is restraint by every independent builder.

Transparency over secrecy. You cannot prove the risk by setting off a sample. Shared belief has to come from open argument. Classified programs invite the thought “they are ahead,” which is how a race starts. Opposite-handed amino acids are not uranium; they are already in ordinary supply chains.

Cooperation over coercion. Sanctions, sabotage, and exclusive clubs tell excluded scientific powers that this is a contest. China is not in NATO. Starting “with friends only” can be destabilizing here.

The four tests have one job: no actor should have both the motive and the ability.

You can now judge a headline policy without memorizing a catalog. Next comes how those tests become two jobs.

The tempting mistake is to treat secrecy and punishment as strength. On this problem they can manufacture the motive they claim to stop.

That is what lets you split the work into talk-with-rivals and raise-the-cost.