DOOMBENCH.COM
Governance and control

OpenAI withholds full GPT-2 weights in a staged-release experiment

OpenAI introduced the 1.5-billion-parameter GPT-2 but initially released only a smaller model because of possible malicious uses. The team, including Ilya Sutskever, framed staged and partnership-based release as an experiment in responsible publication.

0 comments · 0 votesOpen discussion

Public discussion is readable by everyone. Sign in to comment, reply, or vote.

No comments yet. Start the discussion.

Send detailed or private feedback →
CURRENT ASSESSMENT · REVISION 1
AWAY FROM DOOM27confidence 76/100

Why it moved the index

The release restriction was a completed safeguard against scalable disinformation and impersonation risks. Its effect was limited by uncertainty about misuse and independent replication.

AUDIT TRAIL

Assessment history

  1. R1
    Away 27 · confidence 76

    Historical backfill adds a dated model-release decision, exact GPT-2 attribution, and Sutskever's authorship.

    14 Aug 2026
SHARE THE FINDINGS

Share this page

DoomBench social sharing card for OpenAI withholds full GPT-2 weights in a staged-release experiment.

  1. DoomBench assesses “OpenAI withholds full GPT-2 weights in a staged-release experiment” as evidence moving away from doom, with magnitude 27 and confidence 76 out of 100 in the governance and control category.

  2. The DoomBench assessment of “OpenAI withholds full GPT-2 weights in a staged-release experiment” is based on reporting from OpenAI and records the editorial rationale, source quality, attribution, and revision history.

  3. DoomBench summarizes “OpenAI withholds full GPT-2 weights in a staged-release experiment” as follows: OpenAI introduced the 1.5-billion-parameter GPT-2 but initially released only a smaller model because of possible malicious uses. The...

    https://www.doombench.com/news/openai-withholds-full-gpt-2-weights-in-a-staged-release-experiment-2019-02-14