AI safety concerns over scaling biological data

A user says they are hearing pitches that labs will spend heavily on biological data, as they did on coding and cybersecurity. They find that prospect worrying alongside claims that reinforcement learning for coding agents has enabled offensive hacking and made the digital world very insecure, while alignment remains a major unsolved problem. They hope for greater confidence on those safety issues before biological data scales up massively. In a follow-up, they ask who is creating an “alignment data factory” and invite direct messages.