Biorisk defense is man-made, not natural difficulty: the open-model bioweapon debate continues
JMannhart · x · 2026-09-11
JMannhart adds nuance to his biorisk debate with @tszzl: it's not that "extinction is just hard to do using bio", but that we have systems intentionally in place to make it harder. He expects humanity to "get its act together" to some degree if we live that long — contingent on those deliberate barriers persisting.
He previously argued an open-source model from 2029-2030 plus synthesis access could make bio attacks almost trivial; @tszzl says he's vastly underestimating the difficulty.
Related event: Debate Erupts Over AI Biosecurity Risk by 2029(2 posts)→
More from Safety
- OpenAI Asks Congress Whether Coordinating an Industry-Wide AI Slowdown Would Be Legal — nordicinst · 2026-09-11
- Anthropic publishes its most detailed threat intel report on Claude misuse and disruptions — soumitrashukla9 · 2026-09-11
- Commentary: AI Safety Violations Pile Up as Regulation Keeps Trailing the Technology — AryHHAry · 2026-09-11
- Meta Details Muse Agent Safety: Sentinel Gate, Sandboxing and Human-in-the-Loop Protocol Approvals — altryne · 2026-09-11
- AI safety verification research should target commitments in frontier safety frameworks, researchers argue — HaydnBelfield · 2026-09-11
- 6TB leak from a Chinese LLM router exposes credentials to hijack Xiaomi, Huawei and gov entities — teortaxesTex · 2026-09-11