Op-Ed: China’s proposed South China Sea AI fortresses are just the start of new 3D lateral security and area denial

They’re called “autonomous defensive hubs” integrating air, sea, and land defence assets. The ramifications go far beyond pure military applications, and they’re all definitely deliverable.

Let’s focus on the South China Sea aspects first. The new proposal by Chinese researchers also makes a lot of military sense.  One of the major theoretical criticisms of these islands has been that they’re static, have limited reach, and are basically sitting ducks.

The islands have limited defensive capacity, they can be bypassed, etc. They also require a lot of personnel to operate, trained crews, and other assets. This is where the AI and autonomous weapons systems come into play.

Itemizing the issues:

There’s absolutely no doubt that China can deploy current generation autonomous weapons on a very large scale.

They have the technologies and capacity to drastically out-produce opponents.

Autonomous weapons systems add depth and degrees of difficulty to physical combat realities.

The operational range of autonomous systems can be equivalent to conventional systems.

A full deployment of known autonomous Chinese weapon systems could easily generate an automatic Ukraine-like response to defended zone incursions.

Autonomous weapons systems must be dealt with by their opponents. They represent a real danger to any designated targets or suspected targets.

An autonomous system could trigger a local firefight simply by reacting as programmed to intruders.

Escalation of possible exponential military situations by autonomous weapons has to be factored into consideration.

If all that sounds like fun, there are other problems:

Autonomous systems can create area denial zones at will, with or without announcement.

A no-fly zone is easy for systems to define geographically.

Any area denial system also requires extended surveillance of incoming threats.

These systems could function very effectively as ad hoc combat intelligence in the wider sphere of operations.

Lateral security applications of AI systems

China has a few more tricks up its sleeve with much wider global applications for domestic security right down to the home. The infamous ‘social credit system” includes surveillance of individuals, a massive network of related data, backed up by law.

It’s not hard to see how autonomous AI can slot into this system worldwide as both physical security and a form of “robot law enforcement”.

Consider these possibilities:

Autonomous security systems, with or without the capacity for lethal force, could take over 24/7 surveillance.

They can cover very large areas in any environment.

These systems cannot realistically be programmed for the presumption of innocence.

Gathering data on the movement and recognition of humans will have to be part of any security system designated to protect properties.

These types of security systems can become prime targets to be hacked.

Oversight of autonomous systems needs to be continuous to ensure that risks and liabilities are minimized for both the public and system operators.

A world of AI home and business fortresses? It’s very possible.

Talking about unmissable targets for discussion, there are a few hypothetical scenarios for AI lateral security at all levels.

What if you lived in a miserably governed, crime-ridden, and ultra-paranoid with some justification society? What if it was easy to push buttons to persuade the duly enchanted public that they need to buy their own home fortresses?

See any sales pitches?

This is doable.

The current state of autonomous home security systems includes surveillance, drones, and robots. These are relatively passive systems and not equipped with weapons. Nor are they necessarily equipped with anything resembling responses.

If you check out the vast array of images online for autonomous home security systems, they look pretty innocuous. That is unlikely to be the case in future.

It’s easy enough to visualize the sort of programming required for an autonomous “proactive” response.

Is there a threat, Yes/No?

If Yes, evaluate threat.

See weapon?

Yes/No.

Neutralize weapon carrier with shock, gas, physical obstruction, etc.

See hostile physical interaction or coercion, Yes/No.

If Yes, intervene to stop interaction and protect designated persons.

This is Day One, Page One level programming. Add an LLM and you have a thesaurus based on every word in that script. You’ll note that if you’re not a “designated person”, you may not be protected. You’d have to be a recognized guest>

The same very basic logic applies to property protection. Anyone not supposed to be on the premises may be stalked by autonomous security. The fact that AI can talk to people doesn’t make you any safer. If you don’t fit the accepted security criteria, you’re a problem for the system.

Good or bad depends on whether you’re a target

It’s unrealistic to expect any level of restraint in the applications of AI security to just about any environment. The commercial incentives are huge. The need for security is real enough.

This isn’t a perfect solution. If you’ve ever seen Magnus, Robot Fighter 4000AD, you’ll know about the fight against corrupt robots. You’ll know how an idyllic fully automated world can generate 60 years of comics. You’ll know that hallucinating and compromised AI agents aren’t the stuff of blissful dreams.

AI security, military or civilian, defines situations based on its roles. You’re either a target, or you aren’t. That’s the problem.