AI can uncover others’ secrets. Can we resist judging and sharing?: Opinion

What if the price of your groceries depended not only on what you bought, but also on what a retailer’s algorithm had worked out about you?

From Oct 1, the US state of Maryland will ban food retailers from using consumers’ personal data to charge them different prices. Such laws are likely to become increasingly common.

Algorithms can piece together fragments of information scattered across the web to uncover private information about your mental states, activities and locations.

Although such capabilities have been around for years, artificial intelligence has made such inferences far easier to draw and more widely accessible. As Simon Chesterman argues in his June 9 Straits Times commentary, “the next frontier of data protection may be inference protection”.

What privacy laws cannot reach

But privacy laws cannot control the conclusions people draw in their own minds. A cluster of exercise routes involving soldiers on fitness tracking app Strava, for example, may reveal the location of a military facility. Someone asking about job openings and being available during office hours may lead us to conclude that they are unemployed.

A story attributed to novelist William Thackeray offers a more captivating scenario: A priest recounts that the first confession he had heard was about a murder. Moments later, a man joins the group and announces that he was the priest’s first penitent, leading everyone present to an unavoidable conclusion.

In these cases, the inferences are made by individual people, not AI. But this exposes a gap that privacy regulations may never be able to close. Even if laws tightly control how personal information is collected, stored, shared and analysed, they cannot fully control what we work out from the information we encounter. By combining it with what we already know, any of us may arrive at a conclusion that intrudes on someone’s privacy.

Isn’t this pointing out the obvious? What people know and how they draw conclusions cannot possibly be policed. Many inferences are also drawn involuntarily. In Thackeray’s story, the priest’s audience could not help but infer that the man was a murderer. We cannot be responsible for thoughts that arise unexpectedly or conclusions that form before we have time to reflect.

The consequences of this regulatory gap become more pressing as AI plays a larger role in how people encounter and interpret information. The implications are especially relevant in Singapore, which has one of the highest social media penetration rates in the world. Research presented by the Institute of Policy Studies in 2026 found that most respondents used AI chatbots, often to seek practical advice or information. As AI-generated conclusions become a more regular part of how we consume information, privacy cannot depend on regulation alone.

What we do next still matters

While we may not be able to resist every inference that enters our minds, we retain some control over what happens next – look away, withhold judgment and do not forward it.

Suppose you glimpse a hospital appointment notification on a colleague’s phone and infer that they may have a stigmatised medical condition. The initial thought may arise involuntarily, but you can still avert your gaze, resist searching for further clues, remind yourself that your suspicion is inconclusive and refrain from sharing it with others.

We are not only capable of governing what we do with such thoughts, but also have an ethical responsibility to do so. Privacy theorist Carissa Veliz calls these the duties of inattention, restraint, ignorance and silence. In everyday terms, this means choosing not to dwell on the private information of others or assuming it is the truth and passing it on.

The privacy harm that accrues to a victim of a deepfake starkly underscores these duties. We may come across an AI-generated intimate image that appears to show a celebrity, and instinctively assume that it is really them. Each time someone dwells on it, seeks to find out more, believes it to be real or hits the forward button, they aggravate the harm.

The first thought is not the last word

Individual restraint, however, has its limits. Some conclusions form so quickly that we have little chance to ignore or question them. In Thackeray’s story, the audience was caught by surprise and given almost no room for doubt. They could not have reacted in time to distract themselves. Indeed, the conclusion was too sensational to ignore. All the information required for the inference was unexpectedly thrust upon them. There was too little margin for doubt and no need for silence – everyone arrived at the same conclusion on their own. With AI presenting its users with information that leads reasonable people to logical inferences, such cases can only become more common.

Some argue that no one’s right to privacy is violated because others couldn’t help making a deduction. But there are ethical considerations to bear in mind: Preventing concrete harm, fairness and human decency.

Data privacy regulation can control how personal information is collected, analysed and circulated, but it cannot govern every conclusion people draw or every choice they make afterwards. Protecting data privacy therefore also depends on the habits we cultivate as users of information. We may not control the first thought – but we can control the second look, the judgment and the forwarding of it.

Adrian Kwek is an associate professor at the Singapore University of Social Sciences’ College of Interdisciplinary and Experiential Learning.

Source: The Straits Times © SPH Media Limited. Permission required for reproduction.

I spent years in prison. Rehabilitation doesn’t end with counselling: Opinion Court of Appeal and Supreme Court’s appellate division to get new judges