More rapid decision-making.
More interconnected systems.
And, eventually, fewer opportunities for humans to intervene.
That scenario is less dramatic than a robot uprising.
It may also be more plausible.
AI Can Also Save Lives
It is important not to turn this discussion into an argument that AI itself is inherently dangerous.
Artificial intelligence has enormous potential to benefit humanity.
AI can help doctors analyze medical images.
It can assist scientific research.
It can improve weather forecasting.
It can help detect cyber threats.
It can identify patterns in enormous datasets.
In military contexts, AI could potentially help reduce casualties by improving situational awareness, detecting threats earlier or assisting defensive operations.
The United Nations itself has emphasized that AI can produce enormous benefits.
In remarks to the AI Seoul Summit in May 2024, Guterres said AI could save lives, create jobs and foster progress, while simultaneously warning about risks including mass surveillance, disinformation and lethal autonomous weapons.
The issue, therefore, is not simply AI versus humanity.
The real question is how humanity chooses to use AI.
A hammer can build a house or cause an injury.
A nuclear reactor can generate electricity or, in the wrong context, contribute to catastrophic destruction.
Technology is rarely morally neutral in its consequences, but neither is it inherently destined for one outcome.
Governance matters.
Human judgment matters.
Institutions matter.
The Race May Be the Problem
Perhaps the most uncomfortable part of Cameron's warning is that even governments that want responsible AI development could feel pressured to move quickly.
Suppose one country announces that it will impose strict restrictions on autonomous military AI.
A rival country does not.
The first country's military leaders may begin worrying that restraint will produce a strategic disadvantage.
They may argue that their country cannot afford to fall behind.
The result is an arms race driven not by enthusiasm but by fear.
This is precisely the dynamic Cameron described when comparing AI development with the nuclear arms race.
The paradox is that every participant may believe it is acting defensively.
Yet collectively, everyone may be making the world less safe.
That is why international agreements could become increasingly important.
Without common rules, countries may have incentives to prioritize speed over safety.
Regulation Cannot Wait Until After the Crisis
One lesson from the history of dangerous technologies is that regulation introduced after a catastrophe is often too late.
If an autonomous weapons system causes a major international crisis, governments will inevitably ask how it happened.
But by then, the consequences may already be irreversible.
That is why international organizations are calling for guardrails before AI capabilities become deeply integrated into military systems.
The UN has repeatedly called for international frameworks governing military AI.
NATO has developed principles intended to promote responsible military AI.
These efforts do not eliminate the risks.
They demonstrate that governments are beginning to recognize them.
The challenge is turning broad principles into enforceable rules.
What Would Responsible AI Warfare Look Like?
If AI is going to be used in military environments, several safeguards become particularly important.
First, humans need meaningful authority over decisions involving lethal force.
Second, military AI systems need rigorous testing before deployment.
Third, commanders need to understand the limitations of the systems they are using.
Fourth, systems need reliable mechanisms for human intervention and shutdown.
Fifth, countries need procedures for investigating failures.
And sixth, international agreements need to establish boundaries around the most dangerous applications.
NATO's revised AI strategy specifically emphasizes testing, evaluation, verification and validation, as well as mechanisms for dealing with unintended behavior.
Those concepts may sound bureaucratic compared with The Terminator.
But they are precisely the kinds of mechanisms that could prevent real-world systems from behaving in unexpected ways.
The Future May Be Less Like Terminator—and More Complicated
The irony is that today's AI revolution may ultimately prove more complicated than Cameron's original nightmare.
In The Terminator, the enemy is obvious.