Draft. A lawyer has not read this page yet, so it isn't in force. Highlighted passages are gaps for counsel to fill.

Acceptable use policy

Offpeak processes files you already have: it upscales images, removes backgrounds and transcribes audio. These rules apply to everything you upload, everything you produce, and any product you build on the service. If you are unsure, don't send it.

Children

Zero tolerance. You must not use the service for any of the following:

• Child sexual abuse material, in any form, whether real, altered or synthetic.
• Sexualising, exploiting or harming anyone under 18.
• Grooming or sextortion, or any material that supports them.
• Distributing obscene or otherwise illegal material to minors.

We report apparent child sexual abuse material to the authorities and preserve what the law requires. For counsel: the reporting body and duty for our jurisdiction, and whether scanning or hash matching is required or advisable. The service does not scan content today.

Non-consensual intimate imagery

No sexual or nude depiction of a real person without their consent. This covers altering, enhancing, upscaling or cutting out such an image, and removing clothing or backgrounds to create or improve one.

Deepfakes and impersonation

• No use of someone's name, image, voice or likeness without consent in a way that is commercial, defamatory, deceptive, illegal or harmful.
• No preparing or improving source material for a deepfake of a real person made without their consent.
• No identity theft, and no posing as a person or organisation you aren't.
• No passing off altered images or transcripts as authentic evidence or records.

Circumvention

• No bypassing safety guardrails, content filters or other protections, whether ours or a model provider's.
• No using this service as a step to get around the safety systems of another AI service.
• No dodging limits, minimums, launch seat rules or suspensions, including through extra accounts or cards.
• No uploading files built to exploit a decoder, escape a sandbox or reach another customer's data.
• No extracting model weights, or reverse engineering the service or its models.

Harm to people

• No threats, harassment, bullying, or incitement of violence against a person or group.
• No slurs, hate speech or discrimination based on a legally protected characteristic.
• No promoting or facilitating self-harm or suicide.
• No terrorism, violent extremism, human trafficking or sexual violence.
• No gratuitous gore, torture, mutilation or animal abuse.

Deception and fraud

• No scams, fraud, defamation or schemes to deceive.
• No disinformation that could harm public health, public safety, elections or civic participation.
• No spam.

Privacy and rights

• No transcribing a person's speech without the consent the law requires.
• No covert surveillance, tracking or biometric identification of people.
• No processing health, financial or other sensitive personal data unless you have a lawful basis and a signed data processing agreement.
• No infringing copyright, trademark, publicity or privacy rights.

Security and law

• No unauthorised access to systems or data, and no malicious or disruptive activity.
• No using the service to create or improve malware.
• No use that breaks the law, or that creates illegal or regulated goods without authorisation.

Who can't use the service

• Anyone under 18.
• Anyone in a comprehensively sanctioned territory, or on a sanctions list.
• Anyone we have previously banned.

Products built on the service

If your users reach the service through your product, you must bind them to rules at least this strict, act on abuse you learn of, report child sexual abuse material, apply any legally required age checks, and cooperate with enforcement.

Enforcement

Breaches can lead to suspension or termination, blocking of IP addresses or domains, and reports to the authorities. Suspension refuses all requests and cancels queued items. Anything already running finishes.

Reporting

Report abuse to [ABUSE REPORT EMAIL: placeholder].