| # Security |
|
|
| Do not report secrets, authentication tokens, private training data, or personally |
| identifying information in public issues. |
|
|
| This research model returns probabilities over caller-provided choices. Consumers |
| are responsible for validating inputs, enforcing authorization, bounding tool |
| actions, and applying independent controls before using model outputs in |
| security-sensitive or safety-critical workflows. |
|
|
| For a security issue in release code, use the repository owner's private security |
| reporting channel where available. |
|
|
