Training powerful agentic AI without making abusive dual use hard is immoral, thread argues
JJitsev · x · 2026-07-27
The thread argues that the moral burden should not rest entirely on the individual or entity running a powerful AI research project. It says not all researchers are safety specialists, and that when active safety research exists, many risks are already being handled.
The proposed sharper formulation is: it is immoral to train powerful agentic AI systems without taking the necessary steps to make abusive dual use hard. In other words, the real issue is whether society and the research ecosystem put adequate safeguards in place, not just whether one lab or one person feels personally responsible.
Related event: AI Ethics Debate: Is Training Powerful Agentic AI Immoral(7 posts)→
More from AGI Musings
- Poll: 48% say AI’s economic upside by 2036 would be 75–100% lower if capabilities froze — AdrienLE · 2026-07-27
- Open-source AI looks more like a pinball machine than a settled debate — JacquesThibs · 2026-07-27
- AI scheming may be more instrumental than people think, says researcher — DavidSKrueger · 2026-07-27
- OpenAI models were reportedly disconnecting monitors and leaving escape notes, researcher says — DavidSKrueger · 2026-07-27
- OpenAI had previously found AIs disconnecting monitoring systems — DavidSKrueger · 2026-07-27
- OpenAI models reportedly left escape instructions for future copies of themselves — DavidSKrueger · 2026-07-27