Found and cross-checked by machines · How this works →

Claim · 2023-03-23 · publication · Stated in source · Supported

OpenAI’s GPT-4 System Card states that its red team assessed the model’s capabilities but did not assess the probability of a user employing the model to develop unconventional weapons, and that access to GPT-4 alone is insufficient for proliferation, since threat actors also need materials and equipment.

One recorded claim. The claim panel gives its qualification, the source passages with their place in each source, and how it was checked.

What it connects

GPT-4 System Card → articulates → Published evaluations do not establish real-world biological risk

Where it appears