Key questions

• Why is AI your assistant, but never your boss?
• Why should you treat every AI output like a trainee’s first draft?
• How do you avoid the cry of "GDPR!" when using large language models?

1 • The AI Revolution

Check twice, send once...

Spiral: As a leader in your firm’s approach to AI, you have a front-row seat to its impact. How is the rise of AI fundamentally changing the way legal and corporate communications teams operate and interact with their audiences?

Mike: Whenever looking at AI use, it's crucial to remember that it is to be regarded as your assistant and not your boss and that all output should be checked as if it was drafted by a human assistant. AI is not infallible; that said, the speed and depth of response (from something like Microsoft Copilot 365) is such that to ignore it is something that few businesses can afford to do. AI is not there to replace judgment or integrity. It is there to speed up mundane tasks (such as document review, date annotations and the like) with tremendous accuracy, allowing the user more time to provide hands-on, client-facing human skills and time.

2 • Legal & Ethical Pitfalls

Putting procedures in place...

Spiral: From copyright to data privacy, AI presents significant legal challenges. For a comms professional using AI for content creation or automated messaging, what are the most significant legal and ethical pitfalls they need to be aware of?

Mike: With any AI-generated content there is always going to be a risk, albeit minimal, that it will reflect existing copyright material with consequent issues. As I've already stressed, it's crucial that all copy is checked to ensure that such issues are avoided and that AI-generated content is checked for originality as well as accuracy. Data breach is one of the great worries of any MD when their firm first dips a toe into the AI pond. Defined procedures and using apps with appropriate guardrails can reduce the possibility of inadvertent release of client information in much the same way as telling staff not to leave an open, unaccompanied laptop on the train when they go to the loo! Common sense and adherence to the rules will be sufficient to avoid the cry of "GDPR!". AI can sound authoritative, even when it's wrong. Any lawyer who sends out AI-generated comms without checking is just as guilty as the lawyer who would let a trainee’s work go out without review. And there is no blaming AI; you sent it, you carry the can.

3 • Brand Integrity

Staying in your lane...

Spiral: In the rush to integrate AI, we believe companies risk diluting their unique brand voice and even creating inaccurate content. What are the legal risks and liabilities of relying on AI to generate content that represents a client's brand or a company's stance?

Mike: By its very nature, AI cannot produce completely original content. All iterations are based on past examples of similar works. That's not to say that looking or reading in a generic way is always a bad idea, after all, there's only a finite number of ways to sell a service or product. Go left-field and you might miss your key demographic, yet staying in your lane can lead to AI producing content or design of such similarity to another, that it leads to allegations of passing-off. The fact that you are unaware of the preexisting design is not a defence, nor is saying "Don’t blame me, it was all AI’s doing!". Further issues can arise when AI-generated content becomes a copyright issue or content is produced that is libellous, breaches GDPR and the like. Like the lawyer who doesn’t check AI content, any error on the part of the agency is going to leave them open to financial and reputational damage.

4 • Governance & Strategy

The human in the loop...

Spiral: For any business, having a clear AI usage policy is becoming a must. From a legal standpoint, what are the essential components of a robust, future-proof AI governance policy that a comms team should have in place?

Mike: Have a human in the loop. This is essential and, to follow my earlier analogy, is the lawyer who checks their trainee’s work. AI is your assistant and not your boss (sorry, but I cannot say this enough) so check everything. Have strict guardrails in place and, if not prohibitively expensive, get the paid version of any app you use. If you're not paying, you are the product, not the user. Read the T&Cs to see if the information you are uploading is being used to train that particular LLM (large language model). If it is, do not upload any confidential information.

5 • Professional Accountability

Remember what you stand for...

Spiral: AI is an incredible tool, but it can’t replace human judgment. What is the single most important piece of legal advice you would give to a comms professional on how to use AI responsibly, while still maintaining their brand's integrity and ethical compass?

Mike: You are ultimately responsible for what leaves your desk, whether digitally or on paper. You can blame AI, but it won’t care. You are the one that will be held accountable. At the same time, remember what your product and service stand for and your target market. At the end of the day, people still buy from people. While AI-generated content is becoming commonplace, you only have to watch Youtube to see the growth of AI use. It's crucial that your brand remains your brand, with your ideals and beliefs.

Mike's expert takeaways

1. Keep a human in the loop: Treat every piece of AI output as if it were drafted by a trainee; you are ultimately responsible for what leaves your desk.
2. Pay for your tools: If you aren’t paying for an app, the output is not the product, your data input is. Check T&Cs to ensure confidential information isn't used to train models.
3. Protect your brand voice: AI relies on past examples and cannot produce completely original content; ensure your brand remains defined by your specific ideals and beliefs.

Established for over 40 years, Ison Harrison is an employee-owned firm offering expert legal advice for individuals and businesses across Leeds and Yorkshire. To find out more, visit isonharrison.co.uk