Freeze further frontier AI advances until they can be made controllable; pursue safety research and benefits from existing models. [1][2]
New statement. September 29 live broadcast reinforces the conditional frontier pause; no reversal inferred.
Source [1] · Broadcast interview transcript
The Lead with Jake Tapper — Tristan Harris interview
Attributed to: Tristan Harris · CNN
Statement
“There's basically one solution here, which is that you pause until you figure out how to do this right.”
Harris answer following explanation that AI scientists do not know how to make advancing systems reliable, before discussion of existing-model benefits. · View original source ↗
Source context. Harris describes frontier labs spending enormous sums on training toward systems they do not know how to control. He argues existing models can support economic growth during a pause and discusses coordination with China.
Scope
Training increasingly powerful models amid rogue behavior and loss-of-control risks.
Conditions
Pause until reliable control is understood; realize benefits of existing models rather than race onward.
Stated rationale
Harris describes frontier labs spending enormous sums on training toward systems they do not know how to control. He argues existing models can support economic growth during a pause and discusses coordination with China.
Source details
primary · CNN broadcast transcript identifies September 29, 2026 with live broadcast timestamp headings; exact broadcast/event date. Newest located direct pace recommendation. Official CHT press index links this original CNN interview. No later contrary personal recommendation found through October 4 searches.
Discovered 4 Oct 2026.
Published 29 Sep 2026.
Accessed 4 Oct 2026.
Source [2] · Broadcast interview transcript
Anderson Cooper 360 — interview with Tristan Harris
Attributed to: Tristan Harris · CNN
Statement
“…we should freeze the frontier. We should not go further until we know how to make it controllable.”
Harris final answer following [20:45:10], immediately before Cooper says Unlock the Pod door, HAL. · View original source ↗
Source context. Harris distinguishes current economic benefits and safety work from the recursive self-improvement race. The selected recommendation is his own final answer, not his reports of employee letters or other public figures. Jacob Ward is a separate guest.
Scope
Further frontier AI capability development and the race toward uncontrollable superintelligence.
Conditions
Freeze further frontier advances until systems can be made controllable; accelerating safety work and using existing AI are separate.
Stated rationale
Harris distinguishes current economic benefits and safety work from the recursive self-improvement race. The selected recommendation is his own final answer, not his reports of employee letters or other public figures. Jacob Ward is a separate guest.
Source details
primary · CNN broadcast transcript identifies September 24, 2026 and live program times; event/broadcast date is exact. Rush transcript may contain transcription errors. Original CNN transcript independently opened with speaker turns. Direct conditional frontier freeze supports PAUSE; desire to accelerate safety is not ACCELERATE for capability development.
Discovered 4 Oct 2026.
Published 24 Sep 2026.
Accessed 4 Oct 2026.
Editorial assessment. Two independently opened original CNN transcripts provide exact live-broadcast dates and personal speaker turns. Neither corporate activity nor a signed letter substitutes for these recommendations. Deliberate contrary searches included safety acceleration, beneficial AI, newer CHT press and podcast appearances, and October 1–4 interviews; no later reversal located. Rush transcript caveat retained. No public change or human approval.
Assessment reviewed
Earlier position ·
PAUSE
Freeze the frontier until controllable; current-model benefits and safety research remain available. [1]
Initial statement. Freeze the frontier until controllable; current-model benefits and safety research remain available.
Source [1] · Broadcast interview transcript
Anderson Cooper 360 — interview with Tristan Harris
Attributed to: Tristan Harris · CNN
Statement
“…we should freeze the frontier. We should not go further until we know how to make it controllable.”
Harris final answer following [20:45:10], immediately before Cooper says Unlock the Pod door, HAL. · View original source ↗
Source context. Harris distinguishes current economic benefits and safety work from the recursive self-improvement race. The selected recommendation is his own final answer, not his reports of employee letters or other public figures. Jacob Ward is a separate guest.
Scope
Further frontier AI capability development and the race toward uncontrollable superintelligence.
Conditions
Freeze further frontier advances until systems can be made controllable; accelerating safety work and using existing AI are separate.
Stated rationale
Harris distinguishes current economic benefits and safety work from the recursive self-improvement race. The selected recommendation is his own final answer, not his reports of employee letters or other public figures. Jacob Ward is a separate guest.
Source details
primary · CNN broadcast transcript identifies September 24, 2026 and live program times; event/broadcast date is exact. Rush transcript may contain transcription errors. Original CNN transcript independently opened with speaker turns. Direct conditional frontier freeze supports PAUSE; desire to accelerate safety is not ACCELERATE for capability development.
Discovered 4 Oct 2026.
Published 24 Sep 2026.
Accessed 4 Oct 2026.
Editorial assessment. Exact live broadcast date and personal speaker turn; provisional independent review pending.