Any serious discussion about online data safety must include the use of AI tools. More and more white-collar professionals are becoming prompt enthusiasts, yet too few take the privacy risks and potential data exposure seriously.
The contrast is especially striking given how LLM chatbots are evolving into assistants, collaborators, and – at times – even digital pals. We often apply stricter data protection habits when shopping online or navigating websites, yet we tend to let our guard down when engaging with conversational AI. The human-like nature of chatbots fosters a false sense of security that leads to oversharing.
Why Oversharing Happens
There’s something about the way Chat GPT answers requests. The experience is designed to be so engaging, so immersive, that it’s easy to let your guard down. In a study performed on 550 participants, it was evidenced that in compassion/emotional situations, AI was preferred 68% of the time even over a trained crisis responder. Here are a few possible reasons:
- Low-friction interaction: No forms, no warnings, no payment – it feels “safe”. We sometimes forget that chatbots are still internet-connected systems.
- Anthropomorphism: We often treat chatbots like human partners. The conversational, human-like tone creates a false sense of intimacy and trust, which makes us more likely to overshare – even things we’d normally withhold in other digital contexts.
- Emotional projection: Many times, offloading of sensitive data or personal frustrations happens when we are in a highly emotional state. The compassionate, non-judgemental tone of LLMs leads people to consider that they do a better job than a licensed therapist.
In mitigating any risk of data oversharing, the first step is to recognize why we drop our guard with AI chatbots. The next is to decide what our personal limits are, when it comes to data sharing.
Here are some key conscious decisions we can take when engaging with AI tools, especially conversational chatbots:
1. Treat AI Interactions as Public, Not Private
Despite feeling private, conversations may be stored, analyzed, or used for model improvement.
Conscious Decisions:
- “I will assume anything I share could become public at any time, could be seen or stored by unauthorized systems or people.”
Action: remember assumption any time you are about to start a new chat
2. Set Personal Boundaries for What You Will (and Won’t) Share
Just like you’d avoid posting your passport photo online, you need similar mental “rules” for AI chats.
Conscious Decision:
- “I will filter out of my conversations any personal identifiers, secrets, or sensitive business data. I won’t share full names, addresses, account details, or anything that could be pieced together into an identity.”
Action: create a bullet list with types of data you don’t want to share, so that it’s easier to maintain awareness.
3. Understand the Tool’s Data Policy
Different tools handle data differently. Some tools store chats to improve models; others offer private modes.
Conscious Decision:
- “I will check whether my conversations are saved or used for training before entering anything sensitive.”
Action: Look for settings like “Do not train on my data” or “Clear chat history.”
For Chat GPT
Personalization – > Memory – > Deactivate
Improve the model for everyone –> Deactivate
4. Be Skeptical of Answers
AI tools may generate false or misleading information (hallucinations), even when sounding confident.
Conscious Decision:
- “I will double-check critical facts, especially legal, medical, or financial advice.”
Action: Ask for sources that can prove important claims: links to articles, studies, etc. Never base high-stakes decisions solely on AI output without human verification.
5. Stay Aware of Emotional Oversharing
Why: Chatbots may feel comforting, but they don’t understand or care in a human sense. In order to avoid oversharing, it is important to understand how they work.
LLM chatbots like Chat GPT generate answers by predicting the next word in a sentence, based on patterns learned from vast amounts of text. They don’t “think” like humans. They don’t understand meaning or have awareness. They use statistical probability to string together words that are likely to make sense given the input (prompt).
In contrast, human reasoning involves things like memory, goals, beliefs, emotions, and cause-effect understanding. LLMs don’t have any of that. They simulate reasoning by mimicking how humans have written or talked about ideas in the past. While they may sound smart, they don’t know anything or truly understand what they’re saying.
Conscious Decision:
- “If I’m feeling emotional or vulnerable, I’ll be cautious about what I reveal – this is not a therapist.”
Action: Consider keeping a journal (offline) or speaking with a trusted human for personal processing.
6. Use “Sandbox Thinking”
Why: AI tools are best used as digital scratchpads, not private notebooks.
Conscious Decision:
- “I’ll use AI to experiment with ideas, not store them.”
Action: Avoid using chatbots as long-term storage for notes, ideas, or drafts without backing them up securely.
7. Log Out or Use Privacy Modes When Possible
Why: Staying logged in or using your real profile increases the risk of long-term data association.
Conscious Decision:
- “I’ll use incognito/private modes or anonymous profiles when exploring or testing AI tools.”
Action: Learn how you can navigate incognito, depending on your preferred browser. For example, if you are using Microsoft Edge, you can open an InPrivate window by selecting and holding (right-click) the Microsoft Edge logo in the taskbar and then, selecting New InPrivate window.
I hope you enjoyed these tips. In the next feature I will put together an AI safety checklist for software developers and other tech professionals. Stay safe online and think before you prompt!