All legal documents

Acceptable Use Policy

Effective October 7, 2026 · v2026-10-07

Effective October 7, 2026

This Acceptable Use Policy ("AUP") is part of the Cobble Network, Inc. ("Cobble") Customer Agreement & Terms of Service (the "Terms") and applies to everyone who accesses or uses Cobble's websites, APIs, inference endpoints, hosted models, dashboards, and related services (the "Services"), whether directly or through a third-party platform, router, or application that uses the Services. Capitalized terms not defined here have the meanings given in the Terms.

The AUP has two parts. Part A describes content and conduct that is prohibited on the Services. Part B describes fair use of subscription plans: the rules that keep self-service plans affordable for everyone. Violating either part is a material breach of the Terms.

Part A — Prohibited Content and Conduct

A.1 Illegal Activity

You may not use the Services to engage in, promote, or facilitate any activity that is illegal under the laws of the United States or the jurisdiction in which you operate, including:

  • generating, soliciting, or distributing child sexual abuse material (CSAM) or any sexual content involving minors, whether real, fictional, drawn, or synthetic;
  • developing, acquiring, or deploying weapons capable of mass casualties, including chemical, biological, radiological, or nuclear weapons, or providing uplift to persons seeking to do so;
  • creating or distributing malware, ransomware, exploits, or tooling whose primary purpose is unauthorized access to, disruption of, or damage to computer systems;
  • planning or carrying out violence, terrorism, or human trafficking;
  • fraud, phishing, identity theft, financial scams, or the generation of fake reviews, fake credentials, or fraudulent documents;
  • the sale or facilitation of illegal goods or controlled substances;
  • violating export control, sanctions, or anti-money-laundering laws; or
  • any other conduct that violates applicable law.

A.2 Harm to Others

You may not use the Services to:

  • harass, threaten, intimidate, bully, or abuse any person;
  • generate or distribute non-consensual intimate imagery, sexualized depictions of real people without their consent, or content that sexualizes minors in any way;
  • create deceptive synthetic media ("deepfakes") of real people, clone a person's voice, or impersonate a real person or organization without authorization (see Terms Section 20);
  • incite hatred, violence, or discrimination against individuals or groups on the basis of protected characteristics;
  • stalk, dox, or publish private information about any person without consent;
  • target, exploit, or manipulate vulnerable individuals, including through unsupervised mental-health, crisis, or therapeutic applications (see Terms Section 21.4);
  • generate content designed to encourage self-harm or suicide.

A.3 Deception and Manipulation

You may not use the Services to:

  • generate disinformation, election manipulation content, or coordinated inauthentic behavior;
  • misrepresent AI-generated content as human-generated where doing so is deceptive or unlawful, or fail to provide AI-interaction disclosures required by law (see Terms Section 21);
  • operate bots, sockpuppets, or spam campaigns on third-party platforms in violation of those platforms' terms;
  • generate plagiarized content, academic dishonesty materials, or content intended to defeat integrity, moderation, or detection systems.

A.4 Privacy, Data, and Regulated Information

You may not submit to the Services, or use the Services to process:

  • personal information you do not have the legal right to process (see Terms Section 23);
  • biometric identifiers or biometric information except as permitted under Terms Section 22;
  • consumer health data, protected health information, or other regulated health data on standard service tiers (see Terms Sections 24 and 25);
  • data relating to individuals located in the European Union, European Economic Area, or United Kingdom (see Terms Section 14); or
  • payment card data, government identification numbers, or authentication credentials belonging to others, except as strictly necessary and lawful for your own application.

You may not use the Services to build facial-recognition databases, conduct surveillance, or infer sensitive characteristics about individuals without their knowledge and consent.

A.5 Platform Integrity

You may not:

  • attempt to gain unauthorized access to the Services, other customers' accounts or data, or Cobble's infrastructure;
  • conduct unauthorized security testing, adversarial evaluation, jailbreaking, prompt-injection attacks, or other "Red Teaming" against the Services or the models it hosts, except as authorized under Terms Section 16;
  • interfere with, disrupt, or degrade the Services, including through denial-of-service attacks, abusive request patterns, or deliberate exhaustion of shared capacity;
  • circumvent, disable, or probe rate limits, budgets, metering, billing, authentication, geoblocking, or safety systems;
  • scrape, harvest, or bulk-collect data from the Services beyond your authorized usage;
  • reverse engineer the Services or attempt to extract non-public infrastructure, routing, or configuration details;
  • use the Services to train, distill, or improve a competing model or inference service by systematically extracting model outputs for that purpose, where doing so violates the applicable model license; or
  • use stolen payment methods, dispute legitimate charges, or otherwise obtain the Services fraudulently.

A.6 Model Licenses

Every model available through the Services is subject to its own license and usage policy (for example, the Qwen License, the Gemma Terms of Use, the Apache 2.0 or MIT licenses, or a model developer's acceptable use policy). You must comply with the license and usage policy of each model you use. Where a model license is stricter than this AUP, the stricter rule applies. The current list of models and their licenses is published in the Cobble model catalog.

Part B — Fair Use of Subscription Plans

Cobble's self-service plans include a dollar-denominated budget of inference per 5-hour window and per week, together with request, concurrency, context, and API-key limits. These limits are published in the Plan Terms and enforced mechanically; Cobble does not maintain a separate, undisclosed fair-use quota. The following conduct defeats the purpose of per-person pricing and is prohibited (see Terms Section 18.11):

  • No resale. You may not resell, sublicense, rent, or commercialize plan capacity or wallet credit, including by operating a shared, metered, or white-label inference service on top of a self-service plan.
  • No key sharing. You may not share a plan's API keys or account credentials across an organization, team, or group of people in a manner that circumvents per-plan limits. A plan is for one person and that person's own agents and applications.
  • No account stacking. You may not hold more than one active self-service plan per person, or create multiple accounts, payment methods, or email aliases to combine or multiply budgets.
  • No limit evasion. You may not rotate keys, accounts, or payment methods, or otherwise manipulate the metering and reservation systems, to evade a budget, ceiling, concurrency, or key limit.

Organizations that need shared or pooled capacity should contact Cobble about team or enterprise plans.

Part C — How Cobble Enforces This Policy

C.1 What Cobble Monitors

Cobble is an inference infrastructure provider. By default, Cobble does not read, store, or review the content of your prompts or completions, and does not apply content filters or classifiers to request content beyond any safety behavior built into the models themselves. Cobble's abuse detection operates on request metadata (for example, request volume and timing, token counts, error rates, model selection, API-key and account activity, payment signals, and network-level indicators) and on reports received from third parties. See the Privacy Policy for details on what Cobble does and does not retain.

Cobble does not enable content logging to investigate suspected violations, and does not retain content for abuse detection or legal purposes (see Terms Section 7.5). Investigations are conducted using request metadata, account and payment records, and the information contained in a report. If Cobble receives a valid legal order compelling prospective preservation of specific content, it complies only to the extent required and notifies the affected customer unless legally prohibited.

C.2 Reporting Abuse

To report content or conduct that you believe violates this AUP, email hello@cobble.network with the subject line "Abuse report" and as much detail as you can provide, including the date and time, the model or endpoint involved, and any identifying information about the request or account. Cobble reviews reports promptly and will act in accordance with the Terms and applicable law.

C.3 Consequences

Depending on the severity and nature of a violation, Cobble may:

  • issue a warning;
  • throttle, restrict, or filter access to specific models, endpoints, or features;
  • suspend or revoke API keys;
  • suspend or terminate the account, with or without notice, and without refund;
  • preserve and disclose information where required by law or to protect Cobble, its providers, its customers, or the public; and
  • report illegal activity (including suspected CSAM, which Cobble reports to the National Center for Missing & Exploited Children as required by law) to law enforcement.

Where the Services are accessed through a third-party platform or router, Cobble may also notify that platform and may block traffic associated with the violating end user or key.

C.4 Changes

Cobble may update this AUP from time to time as described in Section 27 of the Terms. Continued use of the Services after an update takes effect constitutes acceptance.

Questions about this policy: legal@cobble.network

Terms of ServicePrivacy PolicyData Processing Addendum

Questions? Contact legal@cobble.network