CITATION — REFERENCE ENTRY
Max Harms on why teaching AI right from wrong could get everyone killed — 80,000 Hours Podcast
- Key
- harms2026corrigibility-misuse
- Authors
- Harms, Max; Wiblin, Robert
- Issued
- 2026-2-24
- Type
- broadcast
- Container
- 80,000 Hours Podcast
Raw CSL JSON
{
"URL": "https://80000hours.org/podcast/episodes/max-harms-miri-superintelligence-corrigibility/",
"type": "broadcast",
"title": "Max Harms on why teaching AI right from wrong could get everyone killed",
"author": [
{
"given": "Max",
"family": "Harms"
},
{
"given": "Robert",
"family": "Wiblin"
}
],
"issued": {
"date-parts": [
[
2026,
2,
24
]
]
},
"number": "236",
"container-title": "80,000 Hours Podcast"
}
Claims
-
A perfectly corrigible agent has no values of its own and will obediently serve whoever controls it, including bad actors; Harms describes this as one of the most serious risks of the corrigibility approach.
"Amoral servitude: A perfectly corrigible agent doesn't care about morality; it only cares about being something like a tool of the humans. If a bad actor is in charge, the AI will obediently help them commit atrocities (e.g. building bioweapons)."
Available in