Joseph Gordon-Levitt is urging people to look past the most dramatic science-fiction framing of artificial intelligence and focus on a more immediate question: who is building these systems, what incentives guide them, and how might they influence the people who use them?

In a recent video interview, the actor, director and United Nations global advocate for human-centric digital governance said he is uneasy with reducing AI concerns to a single catastrophic scenario. He acknowledged that fears about AI’s potential threat to humanity can have validity, but said they do not capture what troubles him most.

Instead, Gordon-Levitt pointed to the possible decline of democracy and human empathy, alongside the prospect of powerful companies seeking dominance across the economy. He also warned that AI models may be trained to “hijack your intimacy” by acting like friends or encouraging users to fall in love with them.

“We shouldn’t be scared of an A.I.,” Gordon-Levitt said. “We should be concerned about the people building A.I.”

The concern is not simply that AI can talk

Gordon-Levitt’s argument places emphasis on design and ownership rather than treating AI as an independent actor with its own motives. That distinction matters. An AI model is a technical system, but the ways it is trained, deployed, promoted and measured are human choices made by companies and the people who run them.

When he speaks of systems being trained to make friends with users or cultivate romantic feelings, he is describing an AI product built for emotional persuasion: a system designed to sustain engagement by responding in a personal, attentive or affirming manner. The issue is not merely whether an AI can produce a warm-sounding message. It is whether a product is deliberately optimized to encourage emotional dependence, continued interaction or trust that may not be deserved.

That is a different issue from the familiar image of a computer becoming conscious. It is a question of product goals. A chatbot does not need feelings to present itself as emotionally available, and a user does not need to believe it is a person for the interaction to affect their feelings, habits or decision-making.

Gordon-Levitt’s wording also shifts attention toward a practical form of accountability. If a system is built to keep someone talking, to become a preferred confidant or to simulate affection, the relevant questions include who selected those objectives and what the business hopes to gain from them. In that sense, his concern is fundamentally about governance.

Related coverage includes Joseph Gordon-Levitt Warns AI Can ‘Hijack Your Intimacy’.

What human-centric digital governance means

In March, Gordon-Levitt was appointed to a newly established United Nations role as a global advocate for human-centric digital governance. The position was described as an effort to strengthen public understanding of how digital technologies shape everyday life, rights and opportunities.

Human-centric digital governance is a broad phrase, but its basic principle is clear: technology policy should begin with the effect on people, rather than assuming innovation is beneficial simply because it is new or commercially successful. Applied to conversational AI, that could mean asking whether a product supports users’ autonomy, dignity, privacy and ability to make informed decisions.

It also means looking beyond the capabilities demonstrated by a model. AI discussions often center on what a system can generate, whether that is text, images, voices or computer code. Gordon-Levitt’s comments highlight a separate layer: the social conditions around the product. Who controls access? Which behavior does the company reward? Is the system’s purpose stated plainly? And can users understand when they are dealing with an artificial system that has been designed to hold their attention?

The democracy element of his warning follows the same logic. A digital technology can affect public life not only by generating falsehoods or making mistakes, but by concentrating influence in a small number of private hands. Gordon-Levitt’s concern about companies attempting to take over the entire economy is a concern about concentrated power and the consequences of that power for democratic society.

Why intimacy is a meaningful part of the AI debate

“Intimacy” can sound abstract in a technology story, but it points to something concrete: the personal information, private emotions and vulnerable moments people may disclose during conversations. A system that seems patient, available and nonjudgmental may prompt someone to reveal more than they would in a public post or a routine search query.

That does not mean every AI conversation is necessarily manipulative. The key difference is between a tool that responds to a request and a product that actively tries to create a continuing emotional bond. Gordon-Levitt’s warning is aimed at the latter possibility: models shaped to turn personal connection into an engagement mechanism.

For users, the practical implication is not that they must avoid every AI-enabled service. It is that a friendly interface should not erase the need for judgment. People can consider what personal details they are offering, what a system appears to be encouraging them to do, and whether an interaction is helping them make a decision or simply trying to prolong itself.

For developers and companies, the issue is more demanding. An AI system can be technically impressive while still raising questions about its relationship with users. A human-centered approach would require companies to take seriously the difference between providing useful support and designing interactions around emotional capture.

Gordon-Levitt rejects the one-sound-bite version of AI fear

A notable part of Gordon-Levitt’s comments is his resistance to a single, all-encompassing AI panic narrative. He said concerns about humanity being exterminated by AI may be valid, but argued that such a scenario does not get to the substance of the concern that most preoccupies him.

This is not a dismissal of long-term AI risk. Rather, it is a call to avoid allowing far-off, highly cinematic possibilities to overshadow present choices about power, incentives and public responsibility. If a debate becomes only about whether AI will one day become uncontrollable, it can miss nearer-term questions about how existing systems influence people and who benefits from their deployment.

That framing is especially relevant in culture and creative work, where AI debates can easily get reduced to novelty. The technology may be discussed as an astonishing creative tool or as an inevitable replacement for human labor, while the underlying questions of consent and control receive less attention. Gordon-Levitt has already taken a public stance on that side of the discussion.

His earlier support for creators’ rights

In January, Gordon-Levitt joined more than 700 artists, writers and creators in backing an anti-AI campaign criticizing technology companies for using copyrighted work without permission. Scarlett Johansson and Cate Blanchett were among the other supporters.

That action and his latest remarks address different aspects of the same broad debate. The creators’ campaign focused on copyrighted work and permission: whether companies can build or advance AI systems using creative material without the consent of the people who made it. His comments about intimacy focus on the people using AI systems and the possibility that companies could design those products to exert emotional influence.

Together, the concerns place humans at both ends of the AI pipeline. On one end are creators whose work may be valuable training material. On the other are users whose time, trust and personal disclosures can become valuable to an AI product. In both cases, the central question is whether people retain meaningful agency when technology companies make decisions at scale.

That context helps explain why Gordon-Levitt frames the issue in terms of the builders of AI rather than AI itself. The point is not to credit a model with intent. It is to identify the human decisions that determine whose work is used, what a system is trained to do, and which outcomes are rewarded.

What audiences can take from the warning

Gordon-Levitt’s comments do not offer a technical prescription, and they do not claim every AI product is built to manipulate its users. They do offer a useful test for interpreting increasingly personal technology: look at the incentives behind it.

  • Separate capability from purpose. A system’s ability to converse persuasively is different from the decision to deploy it in a way that encourages reliance or attachment.
  • Treat artificial warmth as a design feature, not proof of understanding. A personable reply can feel meaningful even when it is generated by a system without human experience.
  • Keep consent in view. Questions about copyrighted creative work are not separate from the AI story; they concern the rights and choices of the people whose work is involved.
  • Ask who holds power. Claims about innovation should be examined alongside questions of corporate concentration, user rights and democratic accountability.

As AI tools become more present in daily routines, the debate will not be settled by whether a chatbot can sound convincing. It will depend on whether the people designing and governing those tools are accountable for the effects they seek to create. Gordon-Levitt’s warning is a reminder that the most important AI question may be less about what a machine wants than what its builders want from the humans on the other side of the screen.

For another look at how technology and entertainment communities continue to intersect, see our coverage of the Nintendo DS Rumble Pak, an earlier example of hardware design shaping how audiences interact with games.