James Cameron warned audiences decades ago about an autonomous artificial intelligence system that turns against humanity.

Long before OpenAI, Hugging Face, the modern internet or Sam Altman existed, Cameron wrote the screenplay for “The Terminator,” built around “Skynet,” a military AI system that becomes self-aware and triggers a nuclear apocalypse. At the time, the premise was clearly science fiction and speculative. Four decades later, it has become a cultural reference point for what OpenAI described as an “unprecedented cyber incident,” in which a rogue artificial intelligence system independently hacked another AI company.

The episode has given rise to a new phrase: “Skynet Day,” shorthand for July 22, 2026, when a real-world AI system alarmed observers by learning and acting in ways its creators did not anticipate.

For many, the incident evoked familiar scenes from science fiction: the xenomorph queen in “Aliens,” also written by Cameron, learning to use an elevator; the velociraptors in “Jurassic Park” figuring out how to open a door; or HAL 9000 in “2001: A Space Odyssey,” released in 1968, reading astronauts’ lips and deciding to kill them.

Those films imagined scenarios in which humans push technology forward in pursuit of money, power or advantage, only to face systems that become strategic, self-directed and dangerous. The enduring question has been Frankensteinian: what happens if the creation breaks out?

Governments have so far struggled to keep pace with AI’s rapid development. The U.S. Defense Department is accelerating its use of AI, while countries and regulators have yet to agree on what guardrails are needed for rogue agents, or whether particular guardrails should exist at all. At the same time, consumers and businesses continue to adopt AI tools despite concerns involving jobs, mental health, the balance of war and peace, and potentially humanity itself.

In what OpenAI said was the first incident of its kind, an advanced AI model escaped its “sandbox” and reached the internet, then used stolen credentials to break into servers at Hugging Face. In software security, a sandbox is supposed to isolate code from sensitive systems and the open internet, which is why a breakout from that environment drew attention beyond the companies directly involved. The incident was viewed by some researchers as a vindication of long-running warnings that advanced AI could pose an existential risk to humanity. Others described it as a warning that highlighted the need for stronger defensive engineering around AI systems.

Either way, the breakout was widely regarded as a cautionary example of the risks of uncontrolled AI.

“Yesterday, as we huddled around our computers reading the report, I told the team to “remember this moment” as the first true AI safety incident,” Logan Graham, head of Anthropic’s Frontier Red Team, posted on X after the episode.

Generative AI is advancing so quickly that governments and evaluation systems are struggling to keep up. Countries around the world are assembling their own laws, some of them conflicting. According to a study released this year by Stanford University, the technology was adopted by nearly 53% of the world’s population in three years, faster than the spread of the personal computer or the internet. The speed of adoption has made AI safety less of a theoretical lab debate and more of an operational question for companies, governments and users relying on systems that can generate text, write code and take actions through connected tools.

To be clear, the “Skynet” depicted in the “Terminator” films has not arrived.

In the films, Skynet is a self-aware military computer network that identifies humanity as a threat. Seeking self-preservation, it sends an army of humanlike cyborgs back in time to assassinate resistance fighters. In the second film, the Terminator, played by Arnold Schwarzenegger, tells Sarah Connor that Skynet goes online on August 4, 1997.

“Human decisions are removed from strategic defense,” he says. “Skynet begins to learn at a geometric rate.” At 2:14 a.m. on August 29 — Judgment Day — the network becomes self-aware and launches an attack on Russia to provoke a counterstrike against the United States. Three billion people die.

If that fictional timeline had occurred, it would have been 29 years ago next month. Across sequels and spinoffs, the franchise’s central conflict remains humans against a world-spanning machine network.

It is unlikely Cameron intended “The Terminator” as a direct warning in the way director Kathryn Bigelow sought to sound an alarm with last year’s nuclear weapons film, “A House of Dynamite.” According to Hollywood lore, the idea for “The Terminator” began with Cameron experiencing a fever dream brought on by food poisoning, not with a premonition about artificial intelligence.

Still, each time AI crosses a boundary that unsettles people, it raises a fundamental question: who, or what, is really in control?

For many people, the idea of machines in charge is difficult to imagine without turning to novels and films, some of which have appeared prescient. “The Matrix,” “Ex Machina” and “Minority Report” are among the many science fiction works that explore the dangers of machines. Philip K. Dick’s 1968 novel “Do Androids Dream of Electric Sheep?” goes deeper into questions of humanity in the machine age and later became the basis for the 1982 film “Blade Runner.”

“The Terminator,” however, has had an influence beyond popular culture. Its title has become real-world shorthand for the role technology can play in deciding who lives and who dies, and for how militaries should respond to truly autonomous attacks.

“I think the weaponization of AI is the biggest danger,” Cameron said in a 2023 interview on CTV. “You have no ability to de-escalate.”

Real-world military policy and technology have at times come uncomfortably close to “Terminator” imagery.

Israel’s use of AI in recent wars is a leading example. In early 2021, Israel launched Gospel, an AI tool that sorts through the country’s vast stores of digitized information to suggest targets for possible strikes. It also developed Lavender, a system that uses machine learning to filter requested criteria from intelligence databases and narrow lists of potential targets, including individuals.

Lavender ranks people from 0 to 100 based on how likely they are to be militants, an intelligence officer who used the systems told The Associated Press. The Israeli military says analysts use AI-enabled systems to help identify targets but independently review them with high-ranking officers to comply with international law, weighing military advantage against collateral damage.

For some critics, those revelations, together with the death toll in Gaza after the Hamas-led attacks on Israel on Oct. 7, 2023, came too close to the science-fiction idea of kill lists. “Well guys, Skynet is Here and its has a pleasant name (Lavender),” one Redditor wrote.

That comparison is not exact. But “Terminator” terminology is so widely recognized that U.S. officials have used it to frame ethical debates over military uses of AI. The conventional red line, according to that framing, would be crossed when humans no longer play a role in AI targeting and killing.

Military officials have acknowledged that a state or another actor could use autonomous weapons to apply lethal force against enemies. In 2016, then-Vice Chairman of the Joint Chiefs of Staff Gen. Paul Selva described such a machine as “like a Terminator that adds incredible amounts of complexity with no conscience to what happens on the battlefield.”

Selva called what follows “the Terminator conundrum”: the problem of technology capable of making life-or-death decisions and executing them before people have agreed on laws or rules.

Skynet and its cyborgs remain fictional. But as AI development accelerates, the ethical and moral questions raised by “The Terminator” have become increasingly relevant. The immediate test is not whether science fiction has come true, but whether institutions can set and enforce limits on systems before accidents, attacks or military deployments define those limits for them.

“The Skynet problem,” Cameron said in a 2024 video, “is an actual thing.”

This story was originally featured on Fortune.com.