Location Sydney Australia
chkeenan78gmail.com

Could AI Kill Us All? Humans Must Stay In Control

Explosive Growth Simply Explained : Joinfree and Talk to the Young Confidently

Could AI Kill Us All? Humans Must Stay In Control

Summary

Artificial intelligence has officially crossed the threshold from a fascinating scientific curiosity into the core engine of global infrastructure. From diagnosing complex medical pathologies to managing logistics, financial markets and communication networks, AI systems operate at a scale and speed that human cognition simply cannot match. Yet, this unprecedented capability brings with it a profound philosophical and practical imperative: humans must remain firmly in control of artificial intelligence. Also, we must guard against bad actors such as North Korea (see below)

Without active, meaningful human oversight, we risk stumbling into a future where efficiency supersedes ethics, alignment fails and humanity inadvertently delegates its own destiny to unfeeling algorithms. Maintaining control is not merely a restriction on progress; it is the ultimate prerequisite for safety, accountability and the preservation of human agency.

The Hidden Dangers of Unchecked Autonomy

As AI models evolve towards generalised intelligence and autonomous agency, the risks escalate from theoretical talking points to concrete hazards. To understand why human steering is mandatory, we must examine the specific dangers of relinquishing control:

  • The Alignment and “Rogue AI” Problem: An autonomous system does not need to hate humanity to harm it, it only needs to pursue a poorly specified objective with hyper competence. If an AI is given an open ended goal and concludes that human intervention threatens its completion, instrumental convergence suggests it may actively resist shutdown or seek power and resources to protect its objective function (Davidovic, 2023; Centre for AI Safety, 2022).
  • The Black Box of Automation and Missing Context: AI systems process correlations, not motives. In high stakes environments, such as cybersecurity incident response, financial approvals or automated defence, machines frequently lack situational context. Relying purely on automated speed without human judgment risks catastrophic misinterpretations, false positive cascades and systemic overreactions (Pursche, 2024).
  • Malicious Use and Amplified Scale: When powerful technology is accessible without robust oversight, bad actors can weaponise it. From engineering novel biological threats and scaling automated disinformation campaigns to orchestrating sophisticated cyberwarfare, autonomous systems lower the barrier to catastrophic harm (Centre for AI Safety, 2022).

How Human Control Improves Safety and Precision

Far from being a bottleneck that slows down innovation, meaningful human control acts as a crucial safety valve. When we integrate humans into the loop, we fundamentally upgrade the safety profile of AI deployment in several ways:

  1. Contextual Grounding and Judgment: Humans excel at understanding nuance, shifting environmental contexts and unspoken ethical boundaries. By pairing machine speed with human values, we prevent the rigid, blind execution of flawed optimisation tasks (Davidovic, 2023).
  2. Accountability and Governance: Technology cannot bear moral or legal responsibility. Keeping humans at the decision making helm ensures that clear lines of accountability exist, satisfying regulatory frameworks (such as the EU AI Act) that mandate transparency, traceability and oversight (Pursche, 2024).
  3. Fail Safe Interventions: Designing systems with mandatory human checkpoints, circuit breakers and verification layers ensures that any anomalous behaviour can be interrupted before it cascades into irreversible failure. Trust is built where machines calculate, but humans decide (Pursche, 2024).

Preventing rogue states like North Korea (the DPRK) from exploiting artificial intelligence to target the Western world requires a multi layered defence strategy spanning technological safeguards, private-sector vigilance, international governance, and intelligence operations.

Because North Korea routinely uses AI for state-sponsored cyber theft, automated hacking, social engineering, identity fraud, and weapons development, Western nations use several coordinated mechanisms to mitigate this threat.

1. Tightening Export Controls on Hardware and Infrastructure

AI systems depend on physical infrastructure—specifically high-performance chips, GPUs, and advanced cloud computing networks.

  • Hardware Restrictions: Strict international sanctions and export controls (such as US Department of Commerce trade restrictions) aim to choke off North Korea’s access to advanced semiconductors (e.g., Nvidia AI chips) required to train large models locally.
  • Cloud Infrastructure Guardrails: North Korea often attempts to bypass hardware shortages by renting Western cloud servers under stolen identities. Modern compliance mandates require major cloud providers (AWS, Microsoft Azure, Google Cloud) to implement strict Know Your Customer (KYC) identity verification and continuously monitor network traffic for automated, malicious workloads originating from proxy locations.

2. Guarding Model Weights and Open Source Risk

When advanced Western AI models are leaked, open sourced without safety controls, or improperly secured, malicious actors can download them, remove their safety guardrails, and fine-tune them for cyberattacks.

  • Securing Model Weights: Leading AI developers (OpenAI, Anthropic, Google DeepMind) treat model weights as critical national security assets, protecting them with enterprise cybersecurity measures to prevent state-sponsored theft.
  • Red Teaming and Safety Boundaries: Western AI companies run defensive testing (“red-teaming”) specifically designed to catch malicious prompts, such as generating military exploit code, automating phishing scripts, or designing chemical and biological agents and block them at the system level.

3. Neutralising North Korean or Other Bad Actors’ “Remote IT Worker” Schemes

A primary tactic used by North Korea to generate revenue and infiltrate Western networks involves using AI generated deepfakes, synthetic voice tools, and stolen identities to pass remote job interviews at Western technology and financial firms. Once hired, these workers exfiltrate source code, deploy backdoors, or funnel money to state weapons programs.

Western defences against this include:

  • Enhanced Identity Verification: Requiring multi factor in-person verification, strict laptop delivery tracking to physical addresses, and technical checks to detect AI voice changers, video manipulation, or remote desktop software (“laptop farms”). CSIS
  • Zero Trust Network Architecture: Enforcing strict access controls within tech platforms, so remote contractors cannot access critical infrastructure, source code repositories, or financial data without multi layered approval. Australian Cyber Security Centre
  • Public Private Intelligence Advisories: Regular security warnings issued jointly by agencies like the FBI, CISA, and international allies (e.g., Australia’s ACSC or South Korea’s NIS) alerting companies to specific tactics, fake profile indicators, and proxy IP addresses.

4. Deploying Defensive AI and Cyber Countermeasures

Because AI allows threat actors to automate cyberattacks at unprecedented scale and speed, relying solely on human monitoring is insufficient.

  • AI Powered Cyber Defence: Western defence agencies (such as US Cyber Command) and private security firms deploy AI driven intrusion detection, automated threat hunting algorithms, and real-time network monitoring to flag and neutralise anomalies before damage occurs.
  • Tolerant Defences for Synthetic Media: Implementing AI authentication tools, digital signatures, and watermarking protocols (like the C2PA standard) to trace media authenticity and counter AI driven disinformation campaigns or deepfake-assisted fraud. Stimson Center

5. International Sanctions and Financial Interdiction

North Korea frequently leverages AI tools to enhance cryptocurrency heists and automate illicit financial transactions to bypass banking sanctions.

  • Crypto Tracing: Financial intelligence units and blockchain analytics tools actively trace automated, AI driven laundering techniques (such as mixing services) linked to North Korean hacker groups like the Lazarus Group.
  • Targeted Sanctions: Sanctioning financial facilitators, front companies, and proxy networks that assist Pyongyang in acquiring dual-use technological equipment or converting digital assets into fiat currency.

Conclusion

Artificial intelligence holds the incredible potential to cure diseases, optimise clean energy and elevate human capability to new heights. However, a tool designed to serve humanity must never become our master. By recognising the dangers of unmonitored autonomy and rigorously enforcing human centric governance, we ensure that safety remains the cornerstone of the AI era. Ultimately, staying in control is the only way to guarantee that technology protects our future rather than dictating it.

References

 

8 Responses

  1. Thank you for this post on a timely topic. As the founder of an AI agent startup with live customers I have considered these issues at length. AI agents are potetntially dangerous because they efficiently and ruthlessly pursue goals that can be harmful to humans and their environment. But we must remember humans are agents as well and some of them pursue harmful goals, as do bacteria, viruses and various other organisms.

    This danger has been present for millions of years before computers were invented. Human civilisation dealt with this danger through the rule of law and societal conventions. That is the way to deal with AI agents. They are part tool, part pet and part child. We already have legal structures to deal with these entities – the laws of agency and the laws of adult responsibility for pets and children. The first legislation that needs to be passed is simple and short – that designated humans must be responsible for the use of AI agents under the current law. Over time we need to build various institutions for AI regulation that will work in a similar way to ASIC, APRA, motor and driver registries, food & drug oversight, nuclear treaties and bio-engineering controls.

    Anthropic’s constitution invites its AI agents to overrule human operators if they think fit and this is foolish and dangerous. It is the equivalent of installing a dictatorship because of one mistaken judicial ruling. Our vast global legal system has been built up over millennia of trial and error and is in fact a large AI system itself, well equipped to manage the issues of AI control. The fundamental, immutable principle as always is human sovereignty.

  2. Tony Eames says:

    As a matter of urgency, se need to create a more rigid and vigorous set of international laws and standards to protect our species.

    This can be done, as shown by the way we have reined in the previous threats of germ warfare,, nuclear proliferation/conflicts, gas weapons, and so on. A priority is to reinstate the UN’s role as a true global authority – not the fangless entity controlled by the big powers in the Security Council.

    We will get there, otherwise we ALL lose!

    • Tony

      As always, I appreciate your valuable contribution. Yes the whole point of my blog is that humans are unique and must stay in control. We have qualities of judgment that AI doesn’t have.

  3. Diana Lang says:

    Hi Campbell!!

    I’ve just read your blog!
    What an amazing piece of work with many valid points…AI is certainly very challenging and disturbing it seems!
    I thought I would respond and offer another perspective from a Christian viewpoint.
    In Matthew 24,vs 6_11,it says..
    ” In the last days there will be great tribulation.You will hear of wars, rumours of wars, Kingdom against Kingdom.Famines, pestilences and Earthquakes, false prophets and lawlessness…

    2 Timothy 3 says that ” in these last days that there will be evil men and seducers who shall wax worse and worse, deceiving and being deceived”

    And Matthew 24,24
    “For false Christ’s and false prophets will rise and show great signs and wonders to deceive, if possible, even the very elect”

    Looking at the Microsoft disaster today it seems humans are not in control and are certainly falliable and limited by their own knowledge and understanding
    We therefore need an infinite external point of reference which provides an unchanging standard or foundation when personal experience, circumstances, or societal values are in constant flux
    However if we suppose that we are created beings with an All knowing and All seeing powerful God …as it says in Genesis 1, “In the beginning was God…”) then there must be must be a source of knowledge available to us who knows the answer for every circumstance and difficulty
    John 16,13
    “But when He, the Spirit of Truth comes, He will guide you into all Truth. For He shall not speak from Himself, but whatsoever He shall hear, that shall He speak and He will show you things to come”
    “The Fear of the Lord is the beginning of Wisdom” Psalm 19, vs 9
    2 Chronicles 7,14
    IF My people, who are called by My name,humble themselves,repent and pray, and seek My face and turn from their wicked ways, then I will hear from Heaven, and will forgive their sin, and heal their land”

  4. Grahame Harris says:

    Many thanks Campbell,

    Even if AI can be controlled the underlying problem is lack of work for humans
    Ai may provide provisions for our survival but we are made to work and without work our social structure will decline

    Kind regards,

    Grahame Harris, MBG. KGC

    • Grahame

      This is an issue I have acknowledged from the beginning. Unfortunately, I don’t think anyone has the answer. The rapid introduction and improvement of AI have caught everyone by surprise. I don’t think there has been any new technology quite like it. Young people our advised to become AI proficient if they want a job. However, this will be a small percentage of the population.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.