Perhaps the starting point is to look at the type of AI platforms generally available. At a recent conference about AI use for HR, speakers addressed the generally available platforms and which platform offered the best information.
The speakers stated the following as useful to HR:
- ChatGPT - It is the largest data base and is good for drafting and brainstorming.
- Claude - Good analysis and good for review of long documents.
- Microsoft Copilot - Good integrator that connects to the user organization’s own data.
- Gemini - Good researcher.
- Perplexity - A good fact checker and always shows its sources.
- Notebook LM - The document expert, as if you upload your own documents, you can ask questions about them, making it good for policies and reports.
These speakers also addressed the “prompts” that are useful to get the best information from AI. Their advice was to first let AI know what it’s role is to be in providing their response, i.e., tell them they are an HR director. Second, give AI some background information or framework, as much as practical. Third, tell them the format of exactly what you want AI to produce. Last, give AI the constraints, by telling AI what not to assume.
The conference speakers also noted that AI can be a great language service provider. There are various AI translation programs available that work quite well and quickly. Some programs work better for rare languages.
Editor’s Note: With any use of AI by HR, it is important to note which AI platforms are open, meaning the information is entered into the public realm of information. Sensitive company information and perhaps inquiries should generally not be entered into an open AI system. The reason is that such information then loses its confidential nature, potentially for purposes of attorney-client and work product privileges, and even company trade secrets and other confidential information. The reckless use of company information on an open AI platform creates significant legal exposure to the company. It is probably the first critical issue of AI usage for HR or other company use.
In determining which of the AI platforms is “closed,” an issue is that there are different type platforms even under the above-listed headings, but one starting point is whether the AI tool trains on your data. If such training is off, the remaining evaluation is essentially the same as for any other software, i.e., contractual protections, security certifications, and data-retention policies. It should be noted that many of these same concepts apply to AI research performed by attorneys. Even a lawyer should not put client information into a tool that trains on your data. There are few guidelines available as to what adequate confidentiality and security protections are, but use of the business or enterprise tier of the platform usually eliminates training on your data. The Florida Bar Special Committee on AI Tools and Resources has published a guide that gives some very plain advice: Free AI models may train on your data, so don’t use them for client work. The Guide has a chart that lists confidentiality as an explicit criterion, marking which tools meet it and links to each vendor’s security documentation. If you follow these guidelines, there is still some question of what adequate confidentiality and security protections really mean.
In evaluating existing policies as to AI-related issues, the necessary actions usually start with a clear and practical AI acceptable use policy that empowers employees while protecting the organization from undue risks. A number of questions should be addressed such as: Which platforms are approved?; What settings are required, for example, should learning mode be turned off?; What types of data are off limits?; and, of course, any policy should require employees to verify the accuracy of the data generated by AI.
This article is part of our May 2026 Newsletter.
View the newsletter online
Download the newsletter as a PDF
