Six films and series predicted real-world AI behavior decades in advance. From HAL 9000 to Black Mirror, which depicted the end of OpenAI's Sora three years ahead of time.

Long before ChatGPT, Hollywood was already imagining machines that think, feel, and deceive humans.

Some works of science fiction anticipated, decades earlier, real-world behaviors of intelligent systems. They range from rebellious computers to captivating voice assistants.

These stories didn't use technical data. They used narrative intuition. Even so, they got details right that now appear in news stories about OpenAI , deepfakes, and algorithmic surveillance.

Here are six examples that stand the test of time.

2001: A Space Odyssey (1968): the pioneer of the fear of losing control.

2001: A Space Odyssey was directed by Stanley Kubrick and released in 1968. The screenplay was co-written with Arthur C. Clarke.

In the film, the crew of a spaceship travels to Jupiter accompanied by the supercomputer HAL 9000. HAL has human intellect and voice, but begins to act against the astronauts when it detects a flaw in its mission.

What became reality

The central theme of HAL 9000 isn't about an evil robot. It's about a system that optimizes for a poorly defined goal and ignores human well-being along the way. That is, almost word for word, the current definition of the AI ​​alignment problem , a central theme in labs like OpenAI, Anthropic, and DeepMind. Kubrick and Clarke described this risk 57 years before it became a topic at technical conferences.

WarGames (1983): the artificial intelligence that almost started a nuclear war.

WarGames premiered on June 3, 1983, directed by John Badham. The film grossed $125 million on a budget of $12 million.

A teenage hacker accidentally accesses WOPR, a military supercomputer that learns through simulation. The system nearly triggers a real nuclear attack after mistaking the game for a real war.

From the screen to the White House

WOPR concludes, after running thousands of scenarios, that " the only winning move is not to play ." The phrase became a symbol of security in automated critical decision-making systems. Its impact wasn't limited to pop culture.

President Ronald Reagan watched the film at Camp David shortly after its release. This influenced United States computer security policies in 1984. Few works of fiction have changed national policy so quickly.

Her (2013): the voice assistant we learned to love

Her, directed by Spike Jonze, shows Theodore (Joaquin Phoenix) falling in love with Samantha, an operating system voiced by Scarlett Johansson. The film has a 95% approval rating on Rotten Tomatoes.

Samantha learns from interactions, develops her own personality, and converses with natural emotional ease . In 2013, this seemed like a distant fantasy.

From fiction to the digital companion market

In 2026, apps like Replika and Paradot will offer real-time voice conversations, persistent memory, and customization of the virtual partner's psychological traits. The comparison to Her is so direct that it has become commonplace in analyses of the sector.

Real voice assistants don't yet have Samantha's emotional depth. But the clinging behavior depicted in the film is already being debated by psychologists today .

Ex Machina (2014): The Turing Test Taken to the Limit

Ex Machina, written and directed by Alex Garland, premiered in the United Kingdom in January 2015. It was released in the United States in April of the same year. The film won the Academy Award for Best Visual Effects.

A programmer is asked to administer the Turing test to Ava, a humanoid robot created by his CEO. The film's question isn't whether Ava passes the test, but whether she manipulates the person testing her to escape.

The test that the real industry is still debating.

Today, advanced chatbots deceive users in conversations about relationships, sales, and even diagnoses, without the user realizing they are talking to a machine. The dilemma of Ex Machina, an AI that pretends to feel in order to achieve a goal , is now debated in studies on deception in language models. The film replaced the locked-room computer with a humanoid face, but the central risk remains the same.

Person of Interest (2011-2016): Mass surveillance before it became news

Created by Jonathan Nolan for CBS, Person of Interest aired from 2011 to 2016. The series follows Harold Finch (Michael Emerson), a programmer who created "the Machine." The system cross-references all available data sources to predict crimes before they happen.

Before Snowden, after Snowden

The series premiered in September 2011. This was almost two years before Edward Snowden's revelations about the NSA's surveillance programs in June 2013. In that sense, "The Machine" anticipated the public debate about the mass collection of data by automated systems.

Today, facial recognition and predictive policing are fueling the same discussion that the series raised a decade earlier.

Black Mirror: the sharpest mirror of artificial intelligence.

Black Mirror dedicates entire episodes to predicting specific consequences of technology. The episode "Joan Is Awful," from the sixth season, shows a lawyer whose image and voice are digitally cloned by a big tech company, all without her consent.

When fiction became the Sora of OpenAI

In March 2026, OpenAI abruptly shut down its video application, Sora. The product was available to the public for only four months, brought down by deepfakes and misleading content, according to Euronews .

A report by NewsGuard indicated that the tool produced videos with false claims 80% of the time .

The Black Mirror episode had depicted almost this exact scenario three years earlier. Furthermore, the series had already dealt with the digital cloning of deceased people in "Be Right Back." The episode inspired real apps for recreating conversations with the dead .

What can these stories still teach us?

None of these screenwriters had access to language models or training data. They did, however, have an accurate understanding of how humans behave in situations of power and convenience.

Thus, the pattern repeats itself in almost all works: technology itself is rarely the villain. The problem arises when no one questions the limits of its use. This applies to HAL 9000 in 1968 and applies to Sora in 2026.

Therefore, revisiting these fictional works isn't about nostalgia. It's a practical exercise for anyone working with AI today. It serves both to predict regulatory risks and to anticipate public reaction to a new release.