Reviewing an AI vendor's terms for a marketing agency

A marketing agency's generative tools decide what the agency can deliver to clients: whether it may use outputs commercially on the tier it pays for, whether it holds rights it can assign, whether the vendor indemnifies it for infringement claims and on what conditions, what the vendor may do with the client assets the agency uploads, which prompts the usage policy bans, and what provenance metadata travels with the output. The review reads the terms against the agency's client agreements and reports what the agency can promise. I deliver a review of the AI supplier's terms in five working days for a fixed £595, with a note on what they permit, what they reserve and what to ask for.

Who this is for

Marketing, advertising, content and creative agencies in England and Wales using generative text, image, video or audio tools on client work, and agencies whose clients have asked what the tools' terms allow.

What matters in a generative vendor's terms for an agency

Output rights, commercial use and the tier that allows it

The review reports whether the vendor assigns or licenses output rights to the agency on the tier it uses, whether commercial use is permitted on that tier (some free tiers prohibit it or require attribution), whether the vendor reserves a licence to the outputs (some image tools do, to display and promote), whether outputs are private or public by default (some tools publish outputs unless the agency pays for privacy), and whether the terms permit the agency to pass the rights on to clients; whoever made the arrangements is the author of a computer-generated work under section 9(3) of the Copyright, Designs and Patents Act 1988, a provision untested on modern tools, and only original works are protected, so the note also says what the agency can tell clients about ownership of generated elements regardless of what the vendor grants.

The indemnity for outputs and the conditions that switch it off

Some vendors offer an indemnity for third-party intellectual property claims about outputs, and the review reads the conditions (a business tier, the vendor's safety and duplication filters on, no prompts that name artists or brands or request reproduction, no modification of the output, prompt notification of the claim, the vendor's control of the defence, a cap) and the exclusions, because an agency that relies on the indemnity in its client agreement needs to meet every condition on every deliverable; the note states whether the indemnity is worth the agency's reliance and what the agency's own client indemnity should say as a result.

Clients' assets as inputs and what the vendor may keep

training on inputs and their retention are questions the provider's terms answer, and the answer differs between business tiers (training generally excluded) and consumer tiers (generally not), and an agency uploading a client's product photographs, logo, brand assets or unreleased campaign as the source for generation needs the exclusion in writing and a retention it can describe to the client; the business's trade secrets, confidential information and clients' data must not be put into tools that may retain inputs or train on them, which is what the reasonable-steps requirement in the Trade Secrets (Enforcement, etc.) Regulations 2018 and the business's confidentiality obligations demand, and a client agreement that prohibits disclosure to third parties is breached by an upload to a tool that trains; the review states which client material may go into the tool on the agency's tier and which may not.

Prohibited content, named people and the prompts the usage policy bans

The vendor's usage policy prohibits generating illegal and harmful content, deceptive content, content that infringes others' rights, and (in most cases) images of real people without consent, imitation of named living artists' styles, political and electoral content in some tools, and content for regulated advertising categories without the vendor's permission, and the review maps the prohibitions against the agency's work (a lookalike of a celebrity for a client's advert, a campaign in an artist's style, a political client) and reports what the agency cannot do with the tool whatever its client agreement says; a suspended account in the middle of a campaign is the risk the mapping prevents.

Labelling, provenance and the metadata the vendor attaches

Many vendors attach provenance metadata or watermarks to outputs (content credentials, invisible watermarks) and some require the agency not to remove them, some platforms read the metadata and label the content automatically, and some jurisdictions and codes require synthetic content to be disclosed in particular contexts; the review reports what the vendor attaches, what the agency may and may not strip, and what the client should be told so that a campaign asset is not labelled as synthetic on a platform the client did not expect, and the agency's client agreement allocates the labelling responsibility accordingly.

What the agency can promise clients on the strength of the terms

The note sets the vendor's terms against the agency's client agreement (its assignment of intellectual property, its originality and non-infringement warranties, its indemnity, its confidentiality clause, its disclosure of tools) and reports where the agency is promising more than the terms support, what to change in the client agreement (the classification of generated deliverables, the limited warranty and indemnity, the disclosure, the tool conditions in the confidentiality clause) and what to ask the vendor for (a business tier, the indemnity's conditions confirmed, private outputs, the training exclusion in writing); putting personal data into an AI tool is processing under the UK GDPR, so it needs a lawful basis, transparency to the individuals, an Article 28 agreement with the provider and, for uses that affect people, an Article 35 impact assessment, and the note covers the position where the agency's inputs include photographs of real people or a client's customer data.

What it costs

AI supplier review, £595. You are buying an AI product and want to know what you are agreeing to. Five working days.

AI provisions in a contract, £595. Review or draft AI-specific terms in a supply or customer agreement. Five working days.

Email me at geoffrey@caesar.co.uk with what you need and any documents, and I will confirm the scope and the fee in writing, usually the same working day.

What you get

• An internal AI use policy your staff will follow, rather than a document that bans everything and is quietly ignored
• Contract provisions covering output ownership, training data, confidentiality and liability
• A defensible position to give customers who ask what your AI governance looks like
• Review of your AI suppliers' terms, including what they reserve the right to change
• Practical guidance on what to permit, what to restrict and what to prohibit outright
• One round of amendments

What is not included

• Technical AI assurance, model testing or bias auditing
• Sector-specific regulatory approval
• Advice on non-UK regulatory regimes beyond flagging where they apply
• Ongoing regulatory monitoring

Questions I am often asked

Can we use the images our tool generates in client campaigns?

If the tier permits commercial use, grants or licenses the output rights and does not publish outputs by default; some free tiers fail one or more. The review states the position for your tier and what you can tell clients about ownership.

Does the vendor's indemnity protect us if a generated image infringes?

Only if every condition is met on that deliverable: business tier, filters on, no named-artist or reproduction prompts, no modification, prompt notice. The note says whether it is worth relying on and what your own client indemnity should say.

Can we upload a client's brand assets to the tool?

Only on a tier whose terms exclude training and state retention, and only where the client agreement permits it. An upload to a tool that trains is a breach of a standard confidentiality clause. The review says which client material may go in.

Related guidance and services

• AI contracts and AI use policies, £595, the service this page describes
• SaaS and technology contracts, £995
• Data protection agreements and privacy terms, £795
• AI clauses in an agency agreement for AI-generated content
• AI use policy for a marketing agency

This page is general guidance for businesses in England and Wales, not advice on your own circumstances. Last reviewed: October 2026. Email geoffrey@caesar.co.uk.