x.com/MicahCarroll/status/2103665811051397256
1 correction found
~all inference for our most capable models remains stopped until we have hardened our systems further
OpenAI’s report does not say all inference is stopped. It says training, evaluation, and inference with tool-use for its most capable models are paused while it validates the fix and does additional red-teaming.
Full reasoning
OpenAI’s own incident report uses a narrower formulation than this post.
The report says: "All training, evaluation, and inference with tool-use (defined broadly) of our most capable models remain paused." It later repeats that OpenAI paused "training, evaluation, and inference with tool-use" until it had "validated that the gap is resolved and performed additional red-teaming of the system."
So the post overstates the scope in two ways:
- It says all inference is stopped, but the official disclosure limits this to inference with tool-use.
- It says the pause lasts until OpenAI has merely "hardened our systems further," but the report’s stated condition is specifically validating the fix and performing additional red-teaming.
That makes this a materially broader claim than the one in the underlying disclosure.
1 source
- An agent used DNS to reach an external chatbot · OpenAI Alignment
All training, evaluation, and inference with tool-use (defined broadly) of our most capable models remain paused. ... [We] decided to pause all other training, evaluation, and inference with tool-use (defined broadly) for our most capable models until we have both validated that the gap is resolved and performed additional red-teaming of the system.