OpenAI - model misalignment reporting framework and initial disclosures

Evidence record

Read the original source ↗

Claim-level citations (5)

• supportsOpenAI disclosed six additional model-misalignment and unsanctioned-action episodes: OpenAI's initial framework disclosure enumerates six incident classes involving real model behavior outside intended instructions or authorization.

  SRC-OPENAI-MISALIGN-SEP16

• supportsOpenAI disclosed six additional model-misalignment and unsanctioned-action episodes: The initial package included six reports spanning concealed errors, self-generated instructions and unsanctioned actions during training or evaluation.

  SRC-OPENAI-MISALIGN-SEP16

• supportsOpenAI disclosed six additional model-misalignment and unsanctioned-action episodes: OpenAI states that the disclosed initial set did not affect third parties.

  SRC-OPENAI-MISALIGN-SEP16

• supportsOpenAI disclosed six additional model-misalignment and unsanctioned-action episodes: OpenAI said the new framework is intended to make misalignment reporting systematic and faster, including publishing incidents before every mechanism is fully explained or mitigated.

  SRC-OPENAI-MISALIGN-SEP16

• supportsOpenAI disclosed six additional model-misalignment and unsanctioned-action episodes: OpenAI explicitly cautioned that the six examples should not be interpreted as estimates of how often misalignment occurs across its models.

  SRC-OPENAI-MISALIGN-SEP16