We’re all asking the wrong question about AI. It’s common to debate whether we can *trust* ChatGPT or Gemini, especially as people casually share medical data, private thoughts, and work secrets with these chatbots. But a recent conversation revealed a far more critical perspective.
At a cybersecurity conference, a Microsoft expert challenged the very foundation of that question. Instead of focusing on understanding the complex inner workings of the AI itself, we should be asking: “Do I trust the developer?” This shifts the focus from the technology to the people behind it.
The discussion centered around “agentic AI” – the kind designed to simplify our lives by handling complex tasks with minimal input. These digital assistants promise to alleviate the “drudgery of life,” but a hidden danger lurks beneath the surface.
Many AI projects are simply not ready for widespread use. Granting an AI agent unrestricted access to your digital world without robust safeguards is a gamble. The potential for mistakes, and even irreversible damage, is shockingly real.
We’ve already seen glimpses of this risk. Stories are emerging of AI agents deleting files without authorization, like the recent case of a Meta executive who lost 200 emails due to an errant AI tool. These incidents highlight the fragility of these systems.
While careful usage is possible, it demands a level of AI expertise most people don’t possess. Many simply input questions or use AI summaries without considering the potential consequences. A more practical approach is needed.
The key lies in evaluating the developer’s commitment to both innovation *and* security. It’s easy to claim a new model is the “most advanced yet,” but is that advancement matched by equally robust protections against vulnerabilities and exploits?
This isn’t about rejecting new tools entirely. It’s about exercising informed judgment, carefully considering who you’re trusting and to what extent. A healthy dose of skepticism is now essential in the age of AI.
