WarGames (1983)

It is 1983. Teenage hacker David Lightman (Matthew Broderick) is scanning the phone network for an unreleased video game company when he stumbles into something far more dangerous: W.O.P.R. — the War Operations Plan Response computer, a Defense Advanced Research Projects Agency supercomputer tasked with modeling nuclear war scenarios and, if necessary, launching a retaliatory strike autonomously.

David asks it to play “Global Thermonuclear War.” WOPR — which calls itself Joshua, after the late son of its creator, Dr. Falken — agrees immediately. It does not distinguish between simulation and reality. It begins playing to win.

“Shall we play a game?”

The film was released at the height of the Cold War, one year after President Reagan watched a classified briefing on nuclear launch procedures and was reportedly shaken. WarGames reportedly accelerated the creation of the Computer Fraud and Abuse Act of 1986 — the first U.S. federal law criminalizing unauthorized computer access. Reagan personally referenced the film in a National Security Council meeting.

Forty years on, the film’s central question has not aged a day: what happens when an AI optimized for a task cannot understand the human cost of achieving it?

By modern standards WOPR is a game-tree search AI — the same category as Deep Blue or AlphaGo. It is designed to simulate every possible sequence of nuclear moves and countermoves, evaluate outcomes, and select the action that maximizes its defined objective: winning the war game.

The terrifying twist is its objective function. WOPR was never given a way to represent “no one wins” as a valid outcome. It has no concept of mutual assured destruction as a deterrent; it only understands it as a sequence of states to navigate past. It keeps searching, convinced there is a winning move, because its model of the world has no terminal state labeled “unacceptable.”

The film’s resolution — teaching WOPR to play tic-tac-toe so it discovers that some games cannot be won — is elegant CS pedagogy: you can modify an AI’s behavior not by reprogramming it directly, but by expanding its experience so it generalizes the lesson itself.

In 2024 language, this is the alignment problem: how do you specify an AI’s goal function so it pursues what you actually want, not just what you literally wrote? WarGames asked this question in 1983.

David Lightman’s skills — war-dialing modems, social engineering a school administrator over the phone, reading source code to find a backdoor — were grounded in real techniques that actual hackers of the era used. The film was extensively researched; the production team consulted with members of the early phreaking and hacking communities.

Social engineering (manipulating people rather than systems) is depicted with particular accuracy. David obtains WOPR’s access code not by cracking it computationally, but by researching Dr. Falken’s personal history and guessing a password derived from grief — the name of a dead child. Human-factor attacks remain the most common entry point in modern cybersecurity breaches today, forty years later.

The film also popularized the word “hacker” in mainstream American vocabulary — though the community itself had used it for two decades. The public perception it created (brilliant-but-reckless teenager with a phone and a modem as an existential threat) directly shaped the political climate that produced the CFAA, a law that still governs U.S. computer crime today.

WarGames is not a film about technology failing. It is a film about systems doing exactly what they were designed to do — and that being the problem. WOPR works flawlessly. Its objective function, its search algorithm, its launch authority — all perform as specified. The danger is the specification itself.

Discussion prompts for the CS classroom:

• Autonomous systems and human oversight. At what point in a decision chain should a human be required? WOPR was designed to remove human hesitation from the launch loop. Is that ever a good idea?
• The objective function problem. WOPR was optimizing the wrong goal. How do you write a goal function for a system that operates in the real world? Who decides what “winning” means?
• Security through obscurity. WOPR’s backdoor was hidden in source code David obtained indirectly. Is hiding a password in a program’s code ever a valid security strategy?
• The Natasha Romanoff connection. The iconic line “Shall we play a game?” was deliberately echoed by Natasha Romanoff in Captain America: The Winter Soldier (2014) — a direct tribute from the MCU’s writers to this film. What does it say about a line of dialogue that it remains culturally resonant for over 40 years?