DWA Law
Many of our clients are asking how law firms are actually using AI, and what they should be doing about it on their own productions. We think the best place to start is with our own practice. If we are encouraging writers and producers to be transparent about their use of AI, we should be equally transparent about ours.
How we use AI in our practice
Legal research is one of the most valuable uses. AI allows us to search a very large body of Australian case law and legislation quickly and to identify the authorities most likely to be relevant. Every authority it suggests is then checked against the official report before we rely on it. That step matters, because AI tools can and do suggest cases that do not exist, apparently because they are designed to provide an answer rather than acknowledge uncertainty.
Defamation review is a good example of how this works in practice. Clients often bring us a complete script, manuscript or broadcast transcript dealing with real people and events, sometimes supported by 20 or more documents. A careful read of all of that material can take many hours. Using AI we can produce a first summary of the likely defamatory imputations across the entire body of material in minutes, applying the legal principles we instruct it to use. That allows us to focus our own time on the passages that carry the most risk.
We can then take the analysis a step further. AI can search publicly available material for sources that may support each imputation, which is where any defence of justification (truth) begins, and can identify the claims that do not appear to be supported. This gives clients clearer visibility of their risk at an early stage and a practical list of what is needed to manage it, whether that is further sourcing, changes to the script or a closer legal review of particular scenes. A lawyer reviews every imputation and every source before any advice is given.
We also use AI for drafting. Working from our client’s instructions and our own analysis of the transaction, we can produce a first markup of a contract in around 20 minutes rather than two hours, and prepare first drafts of new agreements from our precedents. This is particularly useful in production, where timeframes are usually tight.
As a general guide AI can assist with roughly the first 60% of a task, being the time-consuming reading, research and drafting. The remaining work, and the judgement, is ours. Up to 40% of an AI-generated first draft is typically revised before it is ready, and preparing a good prompt in the first place requires real skill and experience.
We even use AI to help research articles like this one. The drafting and review, however, are always our own, and every article is written and checked by our lawyers before it is published.
What we never put into AI tools
Confidential or privileged client information is never entered into a publicly available AI tool. The tools we use are configured so that information remains confidential and within the firm. We give our clients the same advice.
How we review AI-assisted work
We review an AI-assisted markup in the same way we would review comments from the lawyer acting for the other party. We read it from start to finish, consider the legal and commercial effect of every change on the production, accept or reject each one and add protective drafting where it is needed. Nothing is sent to a client or another party until a lawyer has completed that review.
For clients the result is contract reviews and advice that are delivered more quickly and often more cost-effectively, with a lawyer remaining responsible for every word.
Our approach
We follow a small number of firm rules.
- Every output is reviewed by a lawyer. No AI-generated work is sent without human review, and the lawyer who sends it is responsible for it.
- We are trained on the tools we use. No one uses an AI tool on client work until they understand how it works and where its limitations lie.
- We work within our professional obligations. Our use of AI follows the Uniform Law conduct rules, the regulators’ joint statement on the use of AI in Australian legal practice and court practice notes such as the Federal Court’s Generative AI Practice Note.
When clients send AI-generated material to their lawyers
There is another side to this. Clients are increasingly running their lawyer’s advice through a chatbot and returning pages of AI-generated questions, or asking their lawyer to review a contract they have drafted with ChatGPT. The Australian Financial Review recently reported on the challenges this is creating for lawyers, including lengthy AI-generated critiques of advice arriving late at night.
An engaged client is a good thing and we understand why clients do this. However reviewing AI-generated material properly takes time, and that time is a cost to the client. It can easily exceed the cost of simply asking the question directly. We will look at this in more detail in a future article. In the meantime we recommend that a law firm’s costs agreement addresses the following.
- Disclosure of AI use. Clients should let their lawyer know when they have used AI to prepare instructions, documents or questions, and which points matter most to them.
- Instructions in the client’s own words. This helps ensure the client’s creative and commercial objectives are clearly understood.
- Confidentiality and privilege. Entering legal advice into a public AI tool may put legal professional privilege and confidentiality at risk, so clients should use only closed or enterprise tools.
- The cost of review. Reviewing AI-generated material is legal work that is charged at the firm’s usual rates and falls outside any estimate, with notice given before any substantial review.
- The firm’s own use of AI. Clients should be told how the firm uses AI and that a lawyer remains responsible for all advice.
If you would like advice on what to include in your own costs agreement or terms of engagement, please contact us.
Why your production needs an AI policy
Most productions now use AI somewhere in their workflow, whether or not it has been formally documented. A writer may use it for research, an editor for a temporary voiceover, the production office for scheduling and a VFX vendor for entire sequences. At the same time financiers, funding bodies, distributors and insurers are increasingly asking which tools were used and how. Screen Australia, for example, lists transparency among its AI guiding principles and has indicated that funding applications will ask about AI use. A short written policy puts a production in a position to answer those questions with confidence.
A well-drafted production AI policy will generally address the following.
- Approved tools. Which AI tools crew and vendors may use, ideally versions that do not retain the production’s material or use it for training.
- Restricted material. Unpublished scripts, personal information of cast and crew, budgets and anything confidential under the production’s financing or distribution agreements.
- Disclosure and approval. Who must disclose AI use, and who approves it.
- Likeness and voice. No digital replica or voice clone of any person without their specific written consent and agreed payment.
- Records. A simple log of what AI was used for and by whom, which will be valuable at delivery and when completing the E&O insurance application.
- Ownership. Contractual terms ensuring that anything created with AI is assigned to the production, and that the production’s material is not used to train any third party’s model.
- Responsibility. A nominated person who maintains the policy and updates it as the law and industry practice develop.
A practical way to implement a policy is to scale approvals to risk. Low-risk administrative tools such as scheduling, budgeting and transcription can be approved by the production manager. Creative assistance such as script breakdowns, concept art or temp tracks should require the producers’ approval. Generative material that will appear in the final deliverable, or anything that uses a real person’s face or voice, should be approved by the producers in consultation with the production’s lawyer. We would also recommend that vendors sign an AI disclosure rider identifying the tools they use, that performer scans are deleted at the end of the project unless otherwise agreed, that AI-assisted pitch material is clearly labelled and that the production’s E&O insurer is told what was used.
Every production is different, and a policy works best when it reflects the project, its financiers and its creative team. If you would like a bespoke AI policy prepared for your production, we would be pleased to discuss it with you.
Disclosure or prohibition?
In our view a blanket prohibition is not the right approach for most productions. It is difficult to enforce and is more likely to discourage transparency than to prevent AI use. Disclosure coupled with approval is generally the more effective approach, although the appropriate level of control depends on the role of each person or service provider.
Writers, directors, composers and designers. Their work forms the creative core of the production, and copyright in that work depends on human authorship. We would recommend requiring disclosure of any AI use and written approval before AI-generated material is included in the deliverable, supported by a warranty that the work is their own.
Performers. This is the area where the industry is furthest from consensus, and where a number of industry bodies are seeking the strongest protections. There has always been some tension between the rights of performers and the rights of producers in the results and proceeds of a performance. AI brings that tension into sharper focus, because a performance can now be reused, altered or recreated well beyond the original shoot.
In Australia the Media, Entertainment & Arts Alliance has called for consent and fair payment whenever a performer’s image, voice or work is used by AI, and has published guidance on protecting performers’ image and work from unauthorised AI use, including AI riders that performers can add to their contracts. Screen Producers Australia has called for consent, reasonable compensation and transparency when screen content is used by AI systems. It is not opposed to AI and wants its members to retain the flexibility to use it in production. There is not yet an industry-wide AI agreement between producers and performers in Australia.
The position in the United States is more developed. SAG-AFTRA’s agreements with the studios require clear consent, a reasonably specific description of the intended use and payment for digital replicas, consent before a performance is digitally altered and bargaining before a synthetic performer replaces a human actor. The 2026 agreement further strengthened the rules on synthetic performers. These terms have not been adopted in Australia, but they provide a useful benchmark.
We act for producers, performers and others across the industry, and our view is a practical one. Any digital replica, voice clone or synthetic version of a real performer should be created only with separate and specific consent and agreed payment, and the contract should clearly set out the uses permitted beyond the original performance. We will compare the US and Australian positions in more detail in a future article. In the meantime this summary of the SAG-AFTRA terms and SAG-AFTRA’s guide to digital replicas for background actors are useful starting points.
Post, VFX and sound vendors. AI tools are increasingly part of these services and prohibiting them may not be practical. We would recommend permitting their use subject to disclosure of the tools used, a clear assignment of outputs and an undertaking that the production’s footage will not be used for training.
Production office and administration. Scheduling, budgeting templates, translation and summaries are generally low risk. Approved tools can be permitted, together with a clear rule about confidential information.
Marketing and publicity. AI can be permitted, subject to approval of anything that uses a cast member’s likeness and with care taken under the Australian Consumer Law to avoid anything that could mislead.
There are productions where a complete prohibition is the right decision. Where a financier, broadcaster or funding body requires it, its requirements should be followed. Documentaries about real people, projects involving archival material or First Nations stories, and productions dealing with sensitive subject matter may also be better served by keeping generative AI out of the creative work entirely.
Whatever approach is chosen, the policy should be in writing, reflected in the production’s contracts and communicated to everyone before work begins.
If you would like assistance preparing an AI policy for your production, please get in touch. You can contact David Whealy at david.whealy@dwalaw.com.au or Brooke Clarke at brooke.clarke@dwalaw.com.au.
Related article
This article is general information only and is not legal advice. Speak to us about your specific circumstances.