In comments on September 15, 2026, a user called for a cost-benefit analysis of specific serious harms enabled by AI models. They argued that the benefits of additional access to biology functions significantly outweighed the risk of easier bioweapon development at that point. In a follow-up reply, they said that balance would likely change, especially if cloud labs became more capable without corresponding improvements in supervision and control. They urged vigilance but were fairly confident existing controls would suffice for a few years.