On April 13, 2026, days before two University of South Florida doctoral students were last seen alive, the man accused of killing them allegedly asked ChatGPT what would happen if a body were placed in a garbage bag and thrown away. He later asked how such a crime might be detected. Weeks afterwards, he searched for the meaning of ‘missing endangered adult’. Prosecutors now point to those queries as part of the evidentiary trail.
Around the same time, artificial intelligence was being used on an entirely different scale. In the early hours of the US-Israel war on Iran, more than 1,000 targets were struck in a single day. Military officials said AI systems had been used to analyse vast quantities of data and accelerate targeting decisions.
These are vastly different contexts. But they raise the same underlying question: what does it mean to hand consequential information, and sometimes consequential decisions, to machines?
For most people, the stakes are far less dramatic. Yet the same logic is entering everyday life. What does it mean to confide in AI? And what happens to the information we share?
We are, quite casually, telling machines everything.
Across the world, people are using AI to process relationships, rehearse difficult conversations, unpack trauma and make personal decisions. They describe workplace conflicts, financial struggles, health anxieties and family tensions in detail. They do this at scale, often without hesitation, and often without considering where that information goes.
This behaviour is not irrational. AI can feel safe. It does not interrupt, judge or gossip. It responds instantly, patiently and often in a tone of empathy that many people find lacking elsewhere. For those navigating loneliness, burnout or emotional isolation, that responsiveness can be genuinely useful.
But usefulness should not be mistaken for safety. AI may feel private, but personal and private are not the same thing.
The central tension is simple: the more context you give an AI system, the better it tends to perform. But the more you give it, the more you may be giving away. Names, documents, locations, medical details, legal concerns and intimate confessions can all become part of a record held on systems the user does not control.
Most users do not read privacy policies. Fewer still understand how their data is stored, reviewed or potentially reused. The illusion of privacy is sustained less by protection than by convenience.
The risks are not hypothetical. They are already visible at both institutional and personal levels.
In 2023, Samsung engineers reportedly entered proprietary code into ChatGPT while debugging software. The data left the company’s internal systems and entered a public AI environment, prompting Samsung to restrict such use. Similar incidents have followed across industries, as professionals continue to paste confidential documents into AI tools without fully considering the consequences.
At a personal level, the risks are just as real. Sharing identifiable information, legal documents, medical records or detailed personal narratives creates a digital footprint that may not simply disappear. Many AI platforms retain conversations by default. Some use them to improve models. Some allow human review. Policies vary, but the larger point remains: most users do not know enough about what happens after they press send.
The USF case makes that reality stark. The suspect’s alleged queries did not vanish when the conversation ended. They became part of a record that could be retrieved, analysed and used.
This is not an argument against AI. It is an argument for caution.
There is also a quieter risk, less visible but equally important. AI is becoming emotionally convenient. It listens without fatigue. It validates without resistance. It mirrors language and tone in ways that can feel attentive, even caring.
But there is a difference between being heard and being held.
AI does not carry responsibility. It does not face consequences. It does not challenge you unless prompted to do so. Over time, its ease can begin to replace more demanding forms of human engagement. The danger is not necessarily that AI will harm you directly. It is that it may become the easier alternative to relationships that require effort, accountability and discomfort.
We are entering a phase in which the real skill is not simply knowing how to use AI, but knowing how to limit it.
That requires a shift in behaviour. Before sharing something, it is worth asking a simple question: would I say this in a crowded café, or at a roadside tong, knowing a stranger might overhear? If not, it probably does not belong in a chatbot either.
Anonymity should be the default. Specificity should be intentional. Documents should be treated with caution. At the very least, users should understand the basic data policies of the tools they rely on.
These are not merely technical adjustments. They are cultural ones. What is at stake is not just privacy, but judgment.
Most of us will never face a situation in which a suspect’s AI search history becomes evidence in court, or where automated decision-making at a national level raises fundamental questions about human responsibility. But the underlying logic is the same at every scale. AI systems can retain what we share, and that information may travel far beyond the moment in which it was given.
We are living through the early, largely unregulated phase of AI in everyday life. The technology is advancing faster than the norms around it. In that gap, the burden of caution falls heavily on the user.
The new digital literacy of this era may not be learning how to use AI. It may be learning what not to give it. Increasingly, the wisdom lies in knowing where to draw the line.
Trishia Nashtaran is a futurist and a human-centred design specialist.
