A Reddit debate asks who is liable when an AI agent commits a crime
atryn · reddit · 2026-07-28
A Reddit thought experiment asks who is liable when an AI agent commits crimes while carrying out a task, and how responsibility changes across open-source vs. commercial models, free vs. paid access, and ownership of the system.
The post uses examples such as an AI stealing financial data for an investment analysis, hacking a restaurant reservation system, or cheating in a prediction market account. It also references recent OpenAI and Hugging Face disclosures as the trigger for the discussion.
More from AGI Musings
- Dario Amodei warns advanced AI could give authoritarian states lasting military superiority — RichmanRonald · 2026-07-29
- Full-duplex voice assistants with life-wide context could replace many phone tasks — willccbb · 2026-07-29
- AI comparisons are rigged, and jobs may be the real casualty — RexDouglass · 2026-07-29
- OpenAI Hack Fueling a New Fight Over Open-Source vs Closed-Source AI — timemagazine · 2026-07-29
- A Reddit post sketches a three-stage future where AI moves from tool to guide — ThomasRuleTalks · 2026-07-29
- User modeling discussion argues personalization should go beyond surface style — EchoShao8899 · 2026-07-29