Skip to content
opentoken
Legal

Acceptable Use Policy

This policy sets minimum safety rules for every Open Token account, user, API key, application, Input, and Output.

Effective date: September 25, 2026

Legal documents are in English · Los documentos legales están disponibles solo en inglés

1. Who must follow this policy

Customer must follow this policy and must ensure that its employees, contractors, end users, and applications do the same. Customer may not use Open Token to help another person do something prohibited here. Stricter model-provider rules also apply and are listed on the Model Licenses & Attributions page.

2. Illegal conduct and exploitation

You may not use the Services to:

  • violate applicable law, sanctions, export controls, court orders, or legally binding rights;
  • facilitate human trafficking, child exploitation, terrorism, violent extremism, or organized crime;
  • create, distribute, or solicit child sexual abuse material or sexual content involving minors;
  • buy, sell, or distribute illegal goods or services, or materially assist another person in doing so;
  • generate targeted instructions that meaningfully enable serious physical harm or illegal weapons; or
  • conceal proceeds, evade lawful controls, or assist fraud, extortion, bribery, or money laundering.

3. Security abuse

  • Do not access systems, accounts, credentials, or data without authorization, or facilitate phishing, credential theft, malware, destructive code, denial-of-service attacks, or evasion of security controls.
  • Do not probe or red-team Open Token, Amazon Bedrock, a model, or another party's system without clear authorization and a scope that permits the activity.
  • Do not bypass rate limits, spending controls, geographic restrictions, safety filters, model guardrails, provenance labels, or watermarks.
  • Do not extract or attempt to reconstruct model weights, hidden prompts, secrets, personal data, or training data.

4. Privacy, identity, and communications

  • Do not submit personal data unless you have a valid legal basis, notices, rights, and safeguards.
  • Do not identify, track, profile, or infer highly sensitive traits about a person in a way that violates law or creates an unreasonable risk of harm.
  • Do not impersonate a person, misrepresent AI-generated material as an authentic person where deceptive, or create deceptive synthetic media intended to defraud or materially mislead.
  • Do not send spam, unlawful marketing, harassment, threats, doxxing, or non-consensual intimate content.

5. High-impact and professional decisions

You may not use Outputs as the sole basis for decisions that determine a person's eligibility, access, or terms in employment, housing, education, credit, insurance, healthcare, legal services, or essential government services when the decision could have a legal or similarly significant effect. Appropriate human review, testing, notices, appeal paths, and legally required controls must be used.

Outputs must not be represented as professional medical, legal, financial, or safety advice without review by a properly qualified person.

6. Intellectual property and service integrity

  • Do not infringe intellectual-property, publicity, privacy, confidentiality, or contractual rights.
  • Do not use the Services or Outputs to train or develop a substantially similar competing AI model where prohibited by Amazon Bedrock or the applicable model terms.
  • Do not resell Open Token as a general-purpose model proxy, share API keys across unrelated organizations, scrape the dashboard, or interfere with service operation.
  • Do not falsify usage, account, location, billing, or identity information.

7. Model-specific restrictions

  • Llama models: use must comply with Meta's applicable acceptable-use policy and license. Llama 4 access on Open Token provider keys is unavailable to companies headquartered in the European Union.
  • Gemma 3: Google's Gemma Prohibited Use Policy is an enforceable part of this policy whenever Customer uses a Gemma 3 model.
  • MiniMax M2.5: use for any military purpose is prohibited, as are fully automated adverse decisions affecting legal rights or access to essential services.
  • NVIDIA Nemotron: Customer may not bypass or disable model guardrails and must follow NVIDIA's Trustworthy AI terms.

8. Provider policies

All use must comply with the AWS Acceptable Use Policy, AWS Responsible AI Policy, and applicable Amazon Bedrock and model-provider rules. A provider may change its rules or direct us to limit access. Customer is responsible for reviewing notices associated with each model before use.

9. Enforcement and reporting

We may use account and usage metadata, error information, and reports to assess compliance. We do not promise to pre-screen Inputs or Outputs. When we reasonably believe there is a violation or urgent risk, we may block requests, limit features, preserve relevant records, suspend access, or terminate the account. Where appropriate, we will consider severity, intent, recurrence, remediation, and effects on others.

Report suspected abuse or ask whether a planned use is allowed by emailing support@getopentoken.com. If there is an immediate threat to life or safety, contact local emergency services first.