The Startup Trying To Save Us From AI Bioweapons

TL;DR

• 1

  A reasoning model found a much better biotech protocol fast: Rajaniemi says Red Queen and OpenAI asked models to improve Gibson assembly, and after two rounds of experiments they landed on a new enzyme-based approach with about a 79-fold to 80-fold improvement over the standard method.

• 2

  The same AI progress that helps cancer vaccines can help bioweapons: Work at Helix Nano on protein design, cancer vaccines, and OpenAI reasoning models made Rajaniemi and co-founder Nikolai Artyomenko realize the exact same tools could lower the cost and expertise needed to design immune-evasive pathogens.

• 3

  Red Queen's thesis is 'new pathogen, press button, get countermeasures': The company wants design time for defenses to approach zero by combining biological reasoning models, antibody design, and rapid biologics manufacturing so outbreaks can be ring-contained the way smallpox and Ebola were.

• 4

  The near-term fear is not sci-fi superplagues but many smaller attacks: Rajaniemi thinks the first wave is more likely to be localized incidents from lone actors, doomsday cults, or school-shooter types who infect dozens or hundreds of people, with success rates rising over time as tools improve.

• 5

  He thinks society needs a broader say in AI than a few labs in San Francisco: Rajaniemi points to citizen assemblies, including examples tied to the LA City Charter and Iceland, as a better model for deciding AI constitutions, regulation, and how much dopamine-optimized AI behavior we should tolerate.

• 6

  His AI philosophy is closer to Frankenstein's parenting lesson than 'caged demon' rhetoric: Rajaniemi argues Mary Shelley's real warning is not 'don't create life' but 'don't abandon what you create,' and says treating AI only as a contained monster may shape it in exactly that direction.