Generative AI and data security: a different kind of talk

127

    The other day I had the opportunity to talk to a coworker about a topic that seems to be in all the headlines: the generative AI and the data security.
    At first glance, it sounds like just another conversation between technologists, full of warnings and legal jargon. But this one was different.

    We're not talking about encryption, firewalls, or regulations. We're talking about people, of trust and how every line of code or prompt we send carries a piece of our digital identity.

    Beyond fear: AI as a reflection of our habits

    Much of the debate surrounding AI security revolves around what things can go wrong: leaks, misuse of information, or loss of control.
    And yes, all of that matters. But what I really took away from that conversation was something deeper:

    “It’s not just a matter of protecting data, but of understanding what data are we providing y because.”

    Every time we ask a model to write, analyze, or create something, we offer them fragments of our thinking, style, and even our way of working.
    In a way, AI doesn't just learn from data: Learn from us.

    From technical security to cultural security

    The real threat is not always technology, but habit.
    We've gotten used to share everything, trusting that the systems will do the right thing. But models don't understand ethics, only patterns.

    Therefore, the conversation about AI safety should focus less on paranoia and more on... collective digital responsibility.
    It's not about distrusting AI, but about using it. with criteria.
    Knowing what information should not leave your work environment, which prompts reveal too much, and when automation stops being convenient and becomes risky.

    The paradox of knowledge

    The more these tools are integrated into our daily lives, the more blurred the line becomes between what is "public" and what is not.
    We worry that AI “knows too much”, but at the same time we constantly feed it with examples, documents and internal contexts.

    There's no magic formula for this. But there is one clear conclusion:

    Safety is not imposed, it is cultivated.
    And it starts with the way each of us understands and manages our own flow of information.


    In summary

    Ultimately, models learn from what we do.
    And perhaps protecting our data starts with to teach the machines the right things.

    ✉ SHALL WE TALK?

    Let's talk

    If we share interests in artificial intelligence, data, or software development, I'd be happy to exchange ideas. You can also contact me for professional matters.

    Read the blog