The basics everyone should understand
AI tools like ChatGPT, Microsoft Copilot, Google Gemini, and Claude are powerful and genuinely useful. They can help you write, summarise information, answer questions, write code, and much more. But they are tools with real limitations and real consequences if misused. Understanding those limitations before you use them is not paranoia — it is practical.
Your data: where it goes
When you type into a commercial AI tool, your input goes to the company's servers, is processed by their AI model, and may be stored. Different tools have different policies about how long they store conversations and whether they use your conversations to train future models. OpenAI (ChatGPT), Microsoft (Copilot), Google (Gemini), and Anthropic (Claude) all publish privacy policies and data settings — it is worth reading the relevant section before using any tool for sensitive matters.
Most major AI tools now allow you to opt out of having your conversations used for training, but opting out is often not the default. Check the settings of any AI tool you use regularly. If you are using an AI tool provided by your employer, ask your IT or legal team what data retention and privacy policies apply.
What not to put into an AI tool
A practical rule: do not put anything into an AI tool that you would not be comfortable having that company's employees potentially see. In practice, this means: no personal information about third parties without their consent — patient names, client details, financial information; no confidential business information unless your employer's AI policy permits it; no passwords, personal identification numbers, or authentication credentials; and no information that could identify individuals in combination even if each piece seems innocuous alone.
AI can be wrong — confidently and convincingly
AI language models can "hallucinate" — generate text that sounds authoritative and accurate but contains invented facts, fake citations, non-existent statistics, and entirely fabricated information. This is not a bug in any single product; it is a characteristic of how large language models work. They predict plausible text, not factually correct text. For any important use — medical questions, legal questions, financial decisions, academic work — verify what an AI tells you against authoritative primary sources. Never publish AI-generated content without reviewing it for accuracy.
Using AI responsibly at work
Many organisations are introducing AI policies that govern what data employees can enter into AI tools and which tools are approved for use. Before using an AI tool for work tasks, check whether your employer has published such a policy. Using an external AI tool with confidential client data, proprietary business information, or regulated data (health records, financial data) may violate your employer's policies, confidentiality obligations, and potentially data protection law — even if you are trying to be more productive.
