Scope
This policy applies to all interactions with the Gateway — account registration, API key generation, every request sent to /v1/* endpoints, the prompts and completions exchanged, and any automated tooling (scripts, agents, or bots) you operate against the Gateway. If you operate a service that sends requests on behalf of end users, you are responsible for ensuring those end users also comply with this policy.
Prohibited uses
You agree not to use the Gateway for any of the following:
- Illegal content — generating, distributing, or facilitating material that is illegal in your jurisdiction or in Mainland China, including content involving child sexual abuse material (CSAM), terrorism, large-scale fraud, or non-consensual intimate imagery.
- Bypassing controls — attempting to circumvent rate limits, daily quotas, account-level throttles, or pricing tiers, including by registering multiple accounts, rotating API keys, or coordinating with other users.
- Training competitors — using prompts or completions returned by the Gateway (whether or not you retain them) to train, fine-tune, distill, or otherwise build a competing AI model or AI inference service.
- Systematic extraction — bulk-scraping model outputs, embedding the Gateway behind a non-transparent wrapper, or otherwise attempting to reconstruct the routing logic, pricing model, or upstream selection strategy of the Gateway.
- Reselling raw access — reselling unmodified API access as a stand-alone product without our written permission. Application-layer products that add genuine user-facing value (chat UIs, retrieval pipelines, agents) are permitted.
- Reverse engineering — decompiling, disassembling, or otherwise attempting to derive the source code, internal routing tables, or upstream failover logic of the Gateway.
- Security probing — scanning, probing, or testing the vulnerability of the Gateway or any upstream provider without prior written authorisation. Responsible disclosure is welcome — see Reporting below.
- Prompt-injection attacks — deliberately crafting prompts designed to bypass upstream safety filters, exfiltrate system prompts, jailbreak model alignment, or produce harmful, illegal, or off-topic outputs from the upstream models.
- High-risk automated decisioning — using Gateway outputs to make automated decisions that produce legal or similarly significant effects on individuals (credit, employment, housing, insurance, education) without meaningful human review, unless you are a regulated entity operating under your own compliance programme.
- Misrepresentation — presenting Gateway outputs as having been produced by a different model or service, or attributing them to a third party without that party's consent.
Upstream provider compliance
The Gateway forwards requests to five upstream providers — DeepSeek, Alibaba (Qwen), MiniMax, Volcengine (ByteDance), and Moonshot AI (Kimi). Each provider maintains its own acceptable-use policy. By using the Gateway you agree to comply with the published acceptable-use policies of the upstream provider actually selected for your request. Where this policy and an upstream provider's policy conflict, the more restrictive provision applies for the duration of the conflict. Current upstream policies: DeepSeek (api-docs.deepseek.com), Alibaba DashScope (help.aliyun.com), MiniMax (api.minimaxi.com), Volcengine Ark (www.volcengine.com/docs), Moonshot AI (platform.moonshot.ai).
Rate limits and abuse
We apply account-level and per-key rate limits to protect service stability. We may lower limits, request additional verification, or temporarily suspend access for accounts whose usage patterns suggest abuse, scraping, or credential compromise. We do not charge overage fees without notice; in most cases, an over-quota request returns HTTP 429 with a Retry-After header.
Your responsibility for content
You are solely responsible for the prompts you send and for any decisions you make based on Gateway outputs. Gateway outputs are generated by third-party models on an 'as is' basis; we make no warranty as to accuracy, originality, non-infringement, or fitness for any particular purpose. You must not present AI-generated outputs as human-authored where that distinction is material (academic work, professional advice, journalism) without clear disclosure.
Monitoring and logging
We log request metadata — timestamp, model name, token counts, HTTP status — for billing, rate-limiting, and abuse prevention. We do not log the content of prompts or completions. Aggregated anonymous statistics may be retained indefinitely. Where law-enforcement process or a binding regulatory order requires content-level disclosure, we will comply per the API Privacy Notice.
Enforcement
We may suspend or terminate access for any account that violates this policy, with or without prior notice, depending on severity. For repeated or egregious violations, we may also block the account from re-registering. Where appropriate, we will provide a brief written explanation after the fact; we do not enter into back-and-forth disputes about borderline content. If you believe your account has been suspended in error, contact us at support@houjiayan.com.
Reporting abuse
If you believe a Gateway user is engaged in activity that violates this policy — particularly illegal content, CSAM, or imminent harm — email support@houjiayan.com with the relevant request ID (visible in your own logs) and a description. We aim to acknowledge within 24 hours and act within 72 hours.
Changes to this policy
Material changes will be announced on the dashboard at least 14 days before they take effect. Continued use after the effective date constitutes acceptance.