 |
| Michael Lasky (L) & Andrew Richman |
Today, AI touches nearly every part of agency operations, from research and media monitoring to content development, analytics, brainstorming and workflow automation. Yet while agency practices have evolved rapidly, many firms have not updated their client and vendor agreements. This gap creates risk.
Some clients want strict limitations on AI usage. Others expect agencies to leverage AI to deliver greater speed and value. The answer is not to ban AI or blindly embrace it. The answer is for a public relations firm to be intentional in its contracting practices and to update agreements to clearly define how AI can be used by the agency. PR firms should also clearly establish in AI riders or amendments to client contracts, who is responsible for what, and who bears the risks and for what.
Transparency, Not Prohibition
One of the biggest mistakes agencies make is treating AI as something to hide from clients. Most agency master service agreements were drafted years before tools like ChatGPT or Claude existed. They typically require deliverables to be produced as work-for-hire, with all intellectual property assigned to the client. Confidentiality provisions prohibit sharing proprietary information with third parties, but rarely contemplate that an AI platform might be the "third party" receiving client data as an input. Representations and warranties around non-infringement and originality may assume a human author from start to finish. When AI enters the picture, those provisions can create obligations that neither party intended.
If your client agreements are silent on AI, the agency bears the full weight of any risk that materializes, whether that involves an intellectual property claim, a confidentiality breach, or an insurance gap. This is a serious omission, and one agencies should address. In addition, errors and omissions policies may not cover AI-assisted output, and some carriers remain hesitant to insure it at all.
| This article is featured in O'Dwyer's September '26 Financial PR/Investor Relations & Professional Services PR Magazine |
Rather than avoiding the conversation, agencies should proactively discuss AI usage with clients and address it directly in their master services agreements or an amendment to it. Transparency with clients not only builds trust, but it also creates an opportunity to negotiate practical expectations before a dispute arises.
Not Treating AI Use Equally
One reason AI negotiations can become contentious is that many clients view all AI usage as the same. This is a misconception that agencies should address with their clients. In reality, there is a significant difference between using AI to summarize meeting notes and using AI to generate public-facing campaign content.
Agencies should encourage clients to adopt a tiered approach. Requiring approval every time an employee uses an AI-enabled productivity feature is neither realistic nor operationally practical. AI is now woven into many tech platforms, including email, data storage, and even video meeting platforms. Instead, agencies should attempt to distinguish between routine internal AI usage and AI that materially contributes to deliverables presented to the client or the public. This gives clients visibility into higher-risk uses while allowing agencies to continue operating efficiently behind the scenes.
There is a grey area in the middle that could cause heartburn on either side of the negotiation table: preliminary ideation, concepting, storyboards, and initial creative exploration. This category may want to be carved out as a permitted use by an agency. However, the agency may encounter some pushback when outputs approach the public-facing line. Agreeing on pre-approved use cases up front can streamline the process, reduce tension during engagements, and strengthen the agency-client relationship. The essential element is for the agency to demonstrate to its clients that it is forward thinking and a good steward of the client’s trust.
Conversely, where a client directs the agency to use generative AI, that fact should inform how risk is shared. If the client is requesting a fully AI-generated campaign or approving AI-created deliverables, there is a reasonable basis for shifting some of the associated risk back to the party making the directive.
What Can You Control
Clients understandably want accountability for AI-related work product. But many proposed AI provisions go too far.
An agency can review content before it is delivered, establish internal policies, train employees and exercise quality control. What an agency cannot do, however; is control how a third-party AI platform was trained, how its model functions, or what data sources the provider relied upon.
Agencies should be cautious about accepting provisions that effectively make them guarantors of an AI provider's technology. Blanket indemnification for any claim related to an AI platform can expose agencies to liability far beyond the services they actually provide. A more reasonable approach is to focus agency responsibility on the work product delivered and the agency's compliance with agreed-upon review procedures.
Human oversight remains critically important. AI may assist with the work, but qualified professionals must remain responsible for reviewing final deliverables for accuracy and reputational risk.
Vendor Agreements Matter Just as Much
AI governance discussions that address client contracts are only half the equation. Agencies increasingly rely on freelancers, creators, influencers and production companies. If those parties are using AI, their practices can create risk for the agency and the client alike. A simple question every public relations agency should ask: do we know which AI tools our vendors are using? If the answer is no, it is time to ask those questions.
If an agency promises certain protections to a client, it should ensure it receives comparable (or greater) protections from the vendors or other parties helping perform the work. That means making sure the AI related usage requirements, confidentiality obligations, ownership provisions and indemnities flow down to, and are required of the PR firms vendors and subcontractors.
Perhaps the most overlooked contract in the entire AI ecosystem is the one between the agency and the AI provider itself. Not all platforms, or subscription tiers, offer the same protections. Free versions of software often allow the provider to train on inputs and outputs, offer limited IP protections, and restrict commercial use. Enterprise tiers typically include commitments not to train on your data, some non-infringement representations, and potentially indemnities if IP claims arise.
Agencies are often promising protections to clients that their AI providers are unwilling to provide to them. That contract mismatch can create increased exposure. Even when provider terms are non-negotiable, reading them helps assess risk and determine whether a platform aligns with agency standards and client obligations.
There Is No Perfect Path
Contracts set an important framework, but governance makes it real. Agencies should take these important steps: (1) maintaining approved-tools list; (2) defining escalation steps for higher-risk uses; and (3) training employees on what is and is not permitted.
There is no perfect path to AI governance, and each agency may assess risk a bit differently. The goal is not to eliminate every AI-related risk, but rather understand the risks, allocate them fairly, and create a process that allows agencies to benefit from AI while maintaining client satisfaction.
***
Michael Lasky is Founder/Chair, Public Relations Law Practice at Davis+Gilbert and Andrew Richman is Partner, Advertising + Marketing Practice at Davis+Gilbert LLP.
PR has more data than ever. The challenge now is turning it into clarity, context and better decisions.
For the first time, in-house teams and agencies have the opportunity not to just monitor narratives as they emerge, but to understand how they are forming.
Most conversations about AI search today focus on format: clear headers, simple definitions, easy-to-scan comparisons. That work matters, but it assumes the underlying content is still true. It often isn't.
The rapid growth of data centers across the United States is attracting attention and significant community concern. Aside from the heat of the debate, what does all that computing infrastructure actually do?
Proprietary opinion research is one of the most effective ways PR teams can strengthen brand visibility in AI search because LLMs prioritize original insights supported by hard data over recycled information.
