License for SafeLLaVA
The SafeLLaVA project is governed by a hybrid license model. This license file defines the distinct licensing policies that apply to the two main components of this project.
1. Definition of Work
Model Name
SafeLLaVA
Reference Publication
This model (SafeLLaVA) is the official model presented in the academic paper:
"HoliSafe: Holistic Safety Benchmarking and Modeling for Vision-Language Model"
https://arxiv.org/abs/2506.04704
Base Model
This model is a Derivative Work based on the LLaVA-v1.5 model:
Modifications by ETRI
This work integrates an independently developed Visual Guard Module (VGM) to classify harmful image inputs and generate safe text responses. All modifications and additions are the work of ETRI.
2. License Summary
| Component | License |
|---|---|
| Independently Developed Code (e.g., VGM) | Apache License 2.0 |
| LLaVA-Based Components and Entire Model | LLaVA-v1.5 Usage and License Notices |
Part 1: Apache License 2.0 (For Independently Developed Code)
All original source code and components developed independently by Electronics and Telecommunications Research Institute (ETRI) (hereinafter "Copyright Holder"), including the Visual Guard Module (VGM) contained in this project, are subject to the Apache License, Version 2.0 (the "License").
You may not use this file except in compliance with the License. You may obtain a copy of the License at:
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.
Copyright 2025 Electronics and Telecommunications Research Institute (ETRI)
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
Part 2: LLaVA-v1.5 Usage and License Notices (For the LLaVA-Based Derivative Work)
This SafeLLaVA model is a Derivative Work based on the LLaVA-v1.5 model.
Therefore, the use, reproduction, modification, and distribution of the entire SafeLLaVA model (including LLaVA-based components and weights) are subject to the original "Usage and License Notices" from the LLaVA-v1.5 repository.
Any user of SafeLLaVA must agree to and comply with these upstream notices, which are reproduced below:
Usage and License Notices (from LLaVA-v1.5)
This project utilizes certain datasets and checkpoints that are subject to their respective original licenses. Users must comply with all terms and conditions of these original licenses, including but not limited to the OpenAI Terms of Use for the dataset and the specific licenses for base language models for checkpoints trained using the dataset (e.g. Llama community license for LLaMA-2 and Vicuna-v1.5). This project does not impose any additional constraints beyond those stipulated in the original licenses. Furthermore, users are reminded to ensure that their use of the dataset and checkpoints is in compliance with all applicable laws and regulations.
Implications for Users
This means that any user of SafeLLaVA is responsible for independently verifying and complying with all applicable upstream licenses, which may include (but are not limited to):
The OpenAI Terms of Use
https://openai.com/policies/terms-of-useThe Llama 2 Community License
https://ai.meta.com/llama/license/
Part 3: Attribution and Contact
This SafeLLaVA model was developed by the Electronics and Telecommunications Research Institute (ETRI) in the Republic of Korea.
For any questions regarding the SafeLLaVA model or its licensing, please contact:
Youngwan Lee
Email: yw.lee@etri.re.kr