AI tools are already a key part of how teams work – for coding, research, customer support, communications, content, and more. But as use grows, so do the legal and operational risks if things aren’t configured properly.
1. Data Protection & Confidentiality
Most GPT tools process what a user inputs in the cloud, so unless settings are managed, those inputs may be stored and used to improve the AI model. Which is fine for generic information and ideas, but not for anything sensitive/ confidential, particularly customer data/ materials/ information, HR information, etc.
Recent reports show that AI transcripts and chat histories have appeared in Google search results (BBC example here). In other words: what feels like a private chat can end up being publicly searchable.
If your team is simply brainstorming generic ideas, that’s fine. But if they’re copying in customer data, personal information, unreleased code, financials, etc., that could be a serious red flag.
What to check:
- Have your team been told not to enter personal data or confidential information into AI tools?
- Do you have an internal AI use policy?
- Have you provided everyone with clear rules and training?
- Have security measures been implemented?
- Are chat histories turned off where appropriate?
- Are you using an enterprise-grade version with clearer data controls?
2. Accuracy and Oversight
GPT can generate fluent, confident content – whether it’s right or wrong. Without review, this could lead to errors in messaging, policy documents, and even client deliverables. Recent examples highlight these risks with BBC research showing accuracy issues in AI-generated answers and a TechCrunch report noting that certain prompts can increase the likelihood of hallucinations.
What to check:
- Are team members relying on AI for business-critical content, important research, legal/ compliance advice, etc, without verification?
- Do you have clear internal guidance/a policy on when GPT output needs human review?
3. Tool Version & Licensing
Not all AI tools are the same. Free versions may lack essential features (like data controls or access to newer models) and be governed by different licensing terms.
What to check:
- Who in your organisation is using AI – and which tools and versions?
- Are there any unapproved browser extensions, plugins, or third-party integrations?
- Do you have central oversight of use, licensing and terms of use?
4. Prompt Discipline = Output Quality
If staff are using GPT prompts differently, you won’t get consistency. Worse, vague/ unclear prompts could lead to inappropriate or incorrect results.
What to check:
- Have you shared prompting best practices or use-case templates?
- Is there a basic usage guide or policy to ensure consistency across teams?
Final Thought
Using AI tools can be like hiring a team of junior assistants – fast, helpful, and scalable. But they need supervision, and often don’t have key information that’s relevant to your business, your legal risks, or your standards – even if you tell them.
A quick health check now could help:
- reduce legal and data risk
- avoid your team’s conversations turning up in Google
- improve team output
- align AI use with your business goals
If you need help checking use and/ or drafting guidance, please get in touch.
