Updates in AI preferences panel

[ffont]

Issue(s)
#2073

Description
This PR incorporates improvements in the Generative AI preferences panel as discussed in the issue above. Main changes are:

• Made a specific Gen AI preferences panel instead of combining the options with the main profile edit
• Added "gen Ai opt-out option"
• Clarified text to explain that the default option is the CC licenses, and that the other options constitute an statement about extra usage restrictions that users request
• Added a special checkbox to override chosen option for the specific case of voice recordings (using BST category solo-speech), and set it to "gen AI opt out". This is in response to specific petitions by users, which wanted extra protection for their voice recordings but might be fine with other types of recordings.
• Improved FAQ text (see below)

What this PR does not include is the possibility of overriding user preferences on a per-sound basis (except for the voice case). If this becomes a necessity, we could add an extra field in the sound description form or use a tag like "no-gen-ai" that will be used to override global settings.

Deployment steps:
This PR needs to run the migrate command after it is deployed.
The relevant text in FAQ should be updated to the one below:

### Can my sounds be used to train Generative Artificial Intelligence (Gen AI) models?

This question is subject of extensive debate, and there is no single unanimous response for it. Nevertheless, in an effort to clarify the situation, as of July 2026 Freesound has introduced a new <a href="/home/ai-preferences/">Generative AI preferences panel</a> through which sound uploaders can make a statement about their preferences regarding the use of their sounds for training Gen AI models. Using this preference panel, <b>additional preferences can be expressed on top of those already defined by the corresponding Creative Commons license applied to each sound</b>. Note that these additional preferences <b>are applied to all uploaded sounds by a specific user</b>, and are independent of the individual license of each sound. The following options are provided:

<ul>
<li>
<b>No additional preferences</b>: Sounds remain subject to the corresponding Creative Commons license terms. Here is a summary of our interpretation of the Creative Commons terms applied to the use case of training AI models:
	<ul>
	<li>CCO sounds have no restrictions.</li>
	<li>CC-BY sounds require disclosure of the training set of the AI model.</li>
	<li>CC-BY-NC sounds require disclosure of the training set of the AI model and the model may not be trained in a commercial setting or used for commercial purposes.</li>
	</ul>
</li>
<li>
<b>Open Source Models only</b>: Sound uploaders express a preference for their sounds to be used for training Gen AI models only when the Creative Commons license terms summarised above are met, and when the trained models are released as Open Source Models and made freely available to the public. Note that selecting this option does not supersede the non-commercial clause of CC-BY-NC sounds, therefore model developers should consider excluding CC-BY-NC sounds when training a model in a commercial setting or if the model is to be used for commercial purposes.
</li>
<li>
<b>Non-Commercial Open Source Models only</b>: Sound uploaders express a preference for their sounds to be used for training Gen AI models only when the Creative Commons license terms summarised above are met, and when the trained models are released as Open Source Models, made freely available to the public, and not trained in a commercial setting nor used for commercial purposes. Note that by using this option, sound uploaders can express a preference against the use of their sounds for training commercial Gen AI models even if the individual sound licenses do not have the non-commercial clause (i.e. for CCO and CC-BY sounds).
</li>
<li>
<b>No generative AI</b>: By selecting this option, sound uploaders express their preference that their uploaded sounds not be used for training any type of generative AI models.
</li>
</ul>

Note that when using the term <i>Open Source Models</i>, we refer to the <a href="https://opensource.org/ai">definition of Open Source AI</a> by the <b>Open Source Initiative (OSI)</b>. In short, that means that for a model to be considered an Open Source Model, it needs to be released under terms that grant the <b>freedom to use, study, modify and share the model</b>, including their code, model weights and training documentation.

The summary of the Creative Commons terms included above is based on our interpretation of how the Creative Commons licenses apply to the generative AI context, which was originally published in the <a href="https://blog.freesound.org/?p=2082"><i>Freesound in the era of generative Artificial Intelligence</i> blog post</a> in June 2024. Please note that uploader's own interpretation might differ from ours.

#### Training Gen AI models with voice recordings 

Complementarily to the preference options above, sound uploaders can make an extra statement to opt-out from Gen AI model training for the particular case of voice recordings. This option is designed to express a preference against the use of such sounds for training models that may generate speech that could be misused to impersonate real people. Note that Creative Commons licenses primarily address copyright-related rights, but voice cloning may involve additional rights, such as personality rights.

The <a href="/home/ai-preferences/">Generative AI preferences panel</a> includes a checkbox that users can tick to express their preference that their sounds classified under the <a href="/help/broad-sound-taxonomy">"Speech &gt; Solo speech" BST category</a> not be used for the purpose of training Gen AI models. If this option is selected, then the <i>No generative AI</i> option above will be applied to all uploaders' sounds classified that category, effectively overriding the global preference.

#### How will my Gen AI preferences be applied?

Third party model developers typically obtain Freesound data either using the <i>Freesound API</i> or by accessing <i>Data Packs</i> that we provide. Here is how the preferences can be applied in each of the cases:

<ul>
<li>
When using the Freesound API, the Gen AI preferences expressed by sound uploaders can be used for <b>filtering sounds</b>. In this way, model developers can take user preferences into account when training AI models. 
</li>
<li>
In order to reduce website strain through web crawlers and ease access for new research, we provide <b>Data Packs for AI training</b> to institutions. These can include universities, research teams, companies. Data Packs include sound data dumps which are ready to be used by model developers, and which contain clear information about uploader preferences and the circumstances under which those preferences apply (in accordance with the Creative Commons license terms and uploaders' Gen AI training preferences). Data Packs are regularly updated, not only including new sounds but also changes in user preferences. 
</li>
</ul>