Content Policy

Please also review the Outpost Code of Conduct for a general overview of our community standards and expectations.

Last updated: August 7, 2024

Outpost's primary aim is to assist the Community in promoting Open, Collaborative, and Responsible Machine Learning. The achievements of Outpost are largely due to the incredible Community on our Platform.

We value these relationships and strive to maintain an environment where individuals feel welcome, supported, and able to make the most of their contributions and experiences. Thus, we have established the following Content Guidelines and Policy (these "Guidelines") for our Platform and its Users.

These Guidelines outline Outpost's efforts to protect our Community. Our objective is to enable the Community to identify and flag types of Content that pose a higher risk of harm, so we can give such Content additional scrutiny as detailed in the following Guidelines.

Please read these Guidelines carefully, as they contain essential information about the types of Content permitted on our Platform.

These Guidelines are incorporated into our Terms of Service, forming a binding agreement between us and you. You should also review our other guidelines, policies, and documents available on our Website, including our Terms of Service and Privacy Policy.

By accessing or using our Platform, you agree to these Guidelines and all other policies and terms. If you do not agree with any of these, please do not use our Platform.

We may update these Guidelines periodically. Changes will become effective 10 days after posting the updated Guidelines on the Website. Continuing to use our Platform after this period signifies acceptance of the changes.

Definitions

Capitalized terms not defined herein shall have the meanings assigned to them in our Terms of Service, Privacy Policy, and all other policies available on our Website.

  • "Content" includes any material posted, displayed, or accessed on our Website or Hub, such as code, data, text, graphics, images, applications, or software provided by you, us, or any third party.

Content types may include:

  • "ML Artifacts": Code and assets hosted in Outpost's Repositories, including Models, Datasets, Spaces;

  • "Community Content": Content found in the Community section of the Outpost Platform, including discussions, comments, usernames, and related documentation such as READMEs, model cards, data cards, pull requests, and merges.

  • "Community" encompasses all Users of the Outpost Platform, including Outpost personnel.

  • "Community Tab" is a collaborative feature for discussing specific Repositories, providing feedback, brainstorming ideas, and opening pull requests for improvements.

  • "Outpost" refers to Outpost Inc., including its affiliates, directors, subsidiaries, contractors, licensors, officers, agents, and employees.

  • "Platform" or "Outpost Hub," or "Hub," refers to the hosting platform where Users can build, benchmark, share, version, and deploy Repositories, which may include Models, Datasets, and Machine Learning Applications.

  • "Repository" is a data structure containing all project files and the entire revision history. Repositories may be:

    • "Public": Visible to anyone on the internet, but only editable by you or members of your organization;
    • "Private": Visible and editable only by you or members of your organization. New Users must join the maintaining organization to see and access its Content;
    • "Gated": Visible to everyone, but access to ML artifacts requires acceptance of conditions in a click-through form or approval by Repository maintainers;
    • "Disabled": Access blocked to all Community members except the owner.
  • "Repository Label" refers to a label assigned to a Repository, such as:

    • "Not For All Audiences" (NFAA): Content determined by Outpost or Repository authors as potentially unsuitable for all Community members, e.g., sexual Content.
  • "Team" or "Outpost Team" refers to Outpost personnel.

Content Assessment and Guidelines

These Guidelines address two categories of Content:

  1. Restricted Content: Inappropriate for the Platform and subject to removal and further consequences based on severity. This Content is detailed in the Restricted Content section and will be handled directly by the Outpost Team.

  2. Moderated Content: Requires an iterative approach to determine appropriate hosting conditions. This Content is detailed in the Moderated Content section and will be addressed in collaboration with the Repository owner and concerned parties, fostering communication and potential improvements to the Repository's code and documentation. We focus on the origin of the ML artifact, its handling by developers, and its usage.

Restricted Content

The following Content is prohibited on our Platform:

  • Unlawful, defamatory, fraudulent, or intentionally deceptive Content, including inauthentic behavior, disinformation, phishing, or scams;
  • Content that harms others;
  • Content promoting discrimination or hate speech (see our Code of Conduct);
  • Harassing, demeaning, or bullying Content;
  • Sexual content used or created for harassment, bullying, or without explicit consent of those represented;
  • All sexual content involving minors;
  • Content promoting or glorifying violence, suffering, or humiliation;
  • Content promoting or inducing unlawful or fraudulent transactions;
  • Content published without explicit consent of those represented;
  • Spam, including product advertising or excessive bulk activity;
  • Cryptomining practices;
  • Content infringing or violating third-party rights or applicable licenses;
  • Content violating third-party privacy;
  • Content violating applicable laws or regulations;
  • Content designed to disrupt, damage, or gain unauthorized access to computer systems or devices;
  • Malware, trojans, viruses, or malicious code;
  • Proxies designed to bypass restrictions by original service providers;
  • Content promoting high-risk activities, including weapons development, self-harm, suicide, gambling, plagiarism, scams, or pseudo-pharmaceuticals.

Content not listed but requiring attention will be addressed on a case-by-case basis.

Additionally, if personal information of individuals under 13 is collected without parental consent, we will remove this data from our Platform (see §7 of our Privacy Policy).

Moderated Content

Moderated Content requires extra scrutiny and a collaborative iterative approach to classify and respond accordingly. This includes Content with a higher risk of harm, such as Content with dual-use potential or uncertain legal status. Community-reported Content triggers the moderation process.

Consent is a core value in this process, supporting forethought and empathy toward stakeholders and proactive measures addressing cultural and contextual factors.

We prioritize collaborative solutions for ML Artifacts and Community Content, involving Repository owners to ensure modifications or additional guardrails meet Guidelines.

ML Artifacts

When Content is reported, the Repository owner typically has 72 hours to respond. Non-engagement or other circumstances may lead to unilateral actions by the Outpost Team. Two levels of intervention are identified:

A. Community features: To reduce problematic outcomes, we may require:

  • Documentation: Publicly request the Repository owner to identify risk factors in the Model or Dataset cards and add the "Not For All Audiences" tag in card metadata.
  • Gating: Publicly request the Repository owner to use the Gated Repository feature.
  • Private: Publicly request the Repository owner to make the artifact private within an Organization.

B. Not-for-all-Audiences tag: Request flagging applicable Content to allow Users to choose whether to see it by default. The Outpost Team may also tag Content meeting specific criteria.

Flagged Content includes, but is not limited to:

  • Unrequested sexual Content: Subjective, cultural, and context-dependent depictions of nudity, partial nudity, sexual topics or themes, pornography, hentai, or ecchi.
  • Unrequested violent Content: Toxic speech, hate speech, trolling, threats, harassment, bullying, targeted misinformation, and disinformation.

C. Outpost tools: If high risk persists despite guardrails, direct actions may be taken:

  • Downgrade: Public discussion on the issue and limiting artifact visibility on the Hub.
  • Private: Public discussion on the issue and making the Repository private.
  • Disable: Public discussion on the issue and disabling the Repository, with only the owner able to access ML artifacts.

Community Content

A collaborative approach is taken to moderate Community Content, ensuring adherence to our Code of Conduct. Measures include:

  • Friendly warning: Private, written warning explaining inappropriate behavior.
  • Consequences: Warning with consequences, such as hiding or closing discussions and restricting Community tab posting for 48 hours.
  • Deactivation: Deactivation from public interaction within the Community Tab.

Reporting Content

Collective responsibility is encouraged to maintain a healthy Community. The Community Tab allows issue discussions, proposing Pull Requests, and suggesting solutions. Harmful Content can be flagged via the "Report" button, opening a public discussion and notifying the Outpost Team.

The Outpost Team may also flag Content based on external requests or concerns. Abusive flagging, such as spamming or harassment, is not tolerated.

Additional Tools

Our Platform offers additional tools for Content moderation:

  • Discussion or pull request authors can edit discussion titles.
  • Repository owners can:
    • Hide community comments;
    • Tag Content as "Not For All Audiences";
    • Gate Repository access, reviewing and approving/rejecting ML artifacts access.

Intellectual Property Rights Infringement

For claims of Content violating intellectual property rights, send a detailed complaint to [email protected], including flagging the allegedly infringing Content. You represent and warrant not providing misleading information.

Contact Us

For feedback, contact us at [email protected] with any questions or concerns.