Notes for commissioners

23 October 2026 · One page for each broadcaster or streamer with published rules, to send with your disclosure · Quotes checked against the documents on 23 October 2026

Each note says what the broadcaster asks, what Shotlore does in those terms, and what you should decide before the shoot. Attach it to your AI disclosure with the security page, the model card and, at delivery, the job's send log. Where a document is only available to partners we say so; always check your own commissioning agreement.

Netflix

Their document: "Using Generative AI in Content Production", August 2025, on the partner portal; the wording below is as reported by trade press, so confirm it with your Netflix contact.

What they ask: tools that "do not store, reuse, or train on production data inputs or outputs"; use "in an enterprise-secured environment" where possible; no Netflix-owned materials ("unreleased assets, scripts, production images") into generative tools without approval; written approval where personal data, talent likeness or third-party IP are involved.

Shotlore in those terms: on a paid Gemini key Google doesn't train on the content and processes it under its DPA, which meets "train" and "reuse"; Google keeps prompts and outputs for 55 days for abuse monitoring, which doesn't literally meet "store". Preview frames of rushes are "production images". On this Mac sends nothing and meets every line.

Decide before the shoot: ask your Netflix contact for written approval for paid Gemini with the 55-day log disclosed, or use On this Mac for this title. Either way, keep the send log and the security sheet with the title's records.

ITV

Their document: "Use of AI in ITV Commissioned Content", version 2, September 2026.

What they ask: an approved-tools list, which includes Gemini; low-risk uses (transcription and translation; production automation such as logging) need no pre-approval and are "notified to ITV as part of the final deliverables"; medium-risk uses, including "any use of AI which involves personal data (such as names, faces, voices, images, or recordings)", need the Commissioning Executive and a Compliance Advisor before filming.

Shotlore in those terms: keywords and captions on previews are logging; transcripts are transcription. Both are low risk in the abstract, but previews that show contributors' faces, and transcripts of their voices, are personal data, which makes them medium risk under ITV's own words.

Decide before the shoot: treat it as medium risk, get the pre-approval, and note in the deliverables that Gemini (an approved tool) was used for logging and transcription, with the send log as the record.

Channel 4

Their document: "AI Editorial Guidelines for Commissioned Content" (4Producers), with the principles PDF of May 2025.

What they ask: "All use of AI in Commissioning needs to be discussed and approved with a Channel 4 Commissioning Editor in advance"; "All AI tools used in production should be documented, including purpose, scope, and human oversight involved"; "No uploading of confidential, personal, or rights-sensitive information into external systems."

Shotlore in those terms: read literally, sending preview frames of unreleased rushes to Gemini is uploading rights-sensitive material into an external system. On this Mac sends nothing and needs only the documentation line.

Decide before the shoot: either use On this Mac for this title, or put paid Gemini to the Commissioning Editor in advance with the security page and the model card, and record the approval. Document the tool, its purpose (cataloguing), scope (previews only; which jobs) and oversight (every suggestion approved by a person) in your AI register.

Sky

Their document: "SKY Framework on using AI in Commissioned Programming".

What they ask: "You must only use secured, self-vetted and 'enterprise' graded tools"; "Never inputting scripts, personal data or any unreleased/owned materials into any public AI"; confirm "that the tool will not reuse, retain, or train on production materials/assets or Sky data"; "SKY must provide written approval before any moderate or high-risk use cases are implemented"; "Maintaining clear auditable records of AI use". The framework names the risk that "Production materials (scripts, footage, metadata) could be stored or cached externally."

Shotlore in those terms: a paid Gemini key is a contracted, enterprise-grade service (Google's DPA) rather than a public chatbot, and Google's terms give the no-training confirmation; the 55-day abuse log is exactly the caching risk Sky names and must be disclosed. The send log is the auditable record.

Decide before the shoot: request Sky's written approval for paid Gemini with the retention stated, or use On this Mac. Keep the send log with the title.

BBC

Their document: Editorial Guidelines, "Guidance: The use of Artificial Intelligence".

What they ask: for independents, "A senior editorial figure … should be responsible and accountable for its use of AI", and material uses should be discussed at commissioning; the guidance warns that "any information input into an AI may be used by the developer of that AI to train it further or be shared with third parties", including "confidential or commercial information, copyright protected content or any personal data". BBC staff may use only tools authorised through the BBC's own software process.

Shotlore in those terms: cataloguing metadata is unlikely to be a "material" impact on content, but the warning about training is answered only by a paid key (no training) and by the on-device route (nothing sent).

Decide before the shoot: name the senior editorial figure, give them this page and Google's terms, raise it at commissioning as a non-material use, and record the decision. If BBC staff will run Shotlore themselves, their own software authorisation applies.

Others

Disney and National Geographic, Apple TV+, Amazon MGM, Warner Bros. Discovery, Paramount, NBCUniversal, CBC, France Télévisions, ARTE, RTVE, SVT, NRK, DR, RTBF and NPO publish no producer-facing AI or supplier-security rules that we could find; they work through the contract, NDAs, partner portals and, for vendors, TPN. Expect a security questionnaire: the security page, the sub-processor list and the configuration guidelines answer most of it. TV 2 Danmark asks that prompts and versions be kept as part of the raw material: the send log does that. Public broadcasters in Germany, Switzerland, Ireland, Belgium, Australia and Canada publish principles that come down to approved tools, no confidential data into public AI, and disclosure; the same three documents serve.

Disclosure wording you can adapt

"Cataloguing of rushes for [title] used Shotlore (Nova Insight Ltd) to suggest keywords, captions and headlines [and transcripts] from preview images [and audio] of the material, which were reviewed and approved by [role] before use. The AI service was Google Gemini ([model]) under the production's own paid Google account: Google does not use the content to improve its products, processes it under its Data Processing Addendum, and deletes it within 55 days. No original files or full-resolution frames left the production's computers. Positions of sensitive locations were rounded or withheld. [Confidential mode was used: no positions, place names, times or speech were sent.] A log of every call is held with the title's records and is available on request."