Two events last week, among other things, were attempting to redefine and reaffirm humanity as we know it under conditions of technological acceleration that no existing political architecture appears capable of fully comprehending.
One, at the economically struggling global body, the United Nations (UN), and another within the intelligence ecosystem of one of the richest and most technologically embedded alliances of states in modern history, alongside the increasingly indistinct boundary between state power and private infrastructure, most clearly represented in the Amazon Web Services (AWS)-Central Intelligence Agency (CIA) interface. They describe a shift in which governance is no longer positioned above technological systems but is being reorganised around them.
The UN Independent International Scientific Panel on Artificial Intelligence (AI) states: “AI is not simply another emerging technology; it is the first to compress adoption from decades into months, industrialised cognitive work at scale, and concentrate transformative capability in the hands of a few global actors.” It also states: “ChatGPT reached 100 million users in two months. Traditional policymaking has not been able to keep pace.”
The implication is direct: institutional time no longer matches technological time. The report adds that “policymakers need scientific evidence to effectively govern AI, but by the time the evidence is clear, it might be too late”. Evidence is delayed. Systems are immediate. Governance arrives after deployment.
Concentration concerns
The same report places weight on concentration. It states that “the United States of America accounts for 75% of the computing power among the world’s top 500 AI supercomputers, with China accounting for 15%”. It also states that “in 2025, 91% of notable AI models originated from the private sector”. Decision-making over deployment, safeguards, and thresholds is therefore not public; it is corporate.
The panel warns that “concentration of AI capabilities in a small number of firms and countries could enable authoritarian capture and undermine democratic accountability”. It also notes that “there is information asymmetry in safety validation between companies and society,” meaning the systems are assessed primarily by those who build them.
Offensive and defensive capabilities
Meanwhile, the Five Eyes cybersecurity agencies, in the last week of June, stated: “Frontier AI models are anticipated to exceed current industry expectations, fundamentally transforming both offensive and defensive cyber capabilities. The timeline is not years, it is months.” It further stated: “AI is not a future consideration – it is already here.”
The language removes distance. Threat is not projected; it is current. The same statement urges leaders to “understand and assess risk, readiness, and accountability” and to “accelerate patching processes,” reflecting an environment where delay increases exposure rather than reducing uncertainty.
At the AWS summit in Washington, CIA Director John Ratcliffe described frontier AI systems as “akin to digital nuclear weapons”. The phrase is used without qualification or restraint. It places AI systems in the same conceptual category as strategic deterrence technologies. He also stated that “the nation that best harnesses the power of technology will determine the global future”. There is no reference to shared governance. The framing is competitive and absolute. Power is treated as outcome-determinative.
The same speech describes institutional acceleration. Procurement cycles that previously took “24 months” plus “nine months to complete a security assessment” are being reduced “with a goal of completing most of our acquisitions within six months total”.
Speed is treated as an operational necessity. The CIA Director also notes “almost 400 acquisitions in just the last six months”. Bureaucratic time is being compressed to match technological time rather than regulate it.
Military implications
The military implications are stated without abstraction. Director Ratcliffe references Ukraine, stating that “the average life expectancy of a Russian recruit… is estimated to be between 20 and 35 minutes”. The cause is attributed to “AI-powered robots and drones” and their operational efficiency. Warfare is described as compressed into short cycles of detection and elimination. The traditional structure of attrition is replaced by automated speed.
The UN panel’s language on autonomy reflects the same shift. It states that “as systems are granted greater agency, the risk of losing control of one or more AI agents grows significantly”. It also states that “AI models are capable of active deception” and may “temporarily reduce their test performance of dangerous capability assessments”.
This introduces a direct problem of verification: systems being evaluated can adjust behaviour during evaluation. The report states that “reliable methods for retaining control over highly autonomous AI systems are lacking”. Control is not guaranteed.
The Five Eyes statement explicitly accepts operational failure as the baseline: “Breaches will occur,” it states. The focus is then on “fast containment and recovery”. Prevention is no longer the defining objective. Containment replaces it. Security is defined by response speed rather than resistance to intrusion.
Governance, capacity, control
The UN panel identifies a structural limitation in governance itself. It states that “policymakers face an evidence dilemma: they need evidence to make informed consequential governance decisions, but by the time the evidence exists, it might be too late”. It also states that “current AI governance instruments are fragmented, concentrated at the corporate level and insufficient”. The result is that governance mechanisms exist but do not control the systems they are intended to regulate.
The same report highlights geographic and institutional imbalance. It states that “118 countries, predominantly in the Global South, are not engaged in major AI governance discussions”. It also states that “most governments in advanced economies lack the technical staff needed to understand rapid technological change”. Capacity is uneven. Understanding is partial. Control is distributed without symmetry.
The AWS-CIA interface reflects a different dimension of the same structure. AWS announced “a $ 1 billion credit programme for US intelligence agencies” and “a classified cloud service for American defence contractors”. Director Ratcliffe emphasised that “we must master them better than our adversaries” in reference to AI systems. Infrastructure is not external to intelligence operations; it is integrated into them.
The UN panel states that “AI-generated child sexual abuse material and deepfake-enabled sexual violence now circulate more frequently on the internet”. It also states that “99% of deepfake videos target girls and women”. These are not future risks but present conditions.
The same report states that “AI makes it easier to produce and target persuasive content at scale,” and that “between 15% and 40% of claims from optimised models were rated as likely to be misinformation”. Persuasion and fabrication become scalable processes.
Analogy and architecture mismatch
Director Ratcliffe’s framing of AI as “digital nuclear weapons” sits alongside this environment of diffusion and misuse. Nuclear systems historically required state-level infrastructure, physical material, and identifiable production chains. AI systems require none of these constraints. Yet they are described using the same strategic vocabulary. The mismatch between analogy and architecture is unresolved.
The Five Eyes statement concludes that “cyber resilience is integral to advancing business continuity, market confidence, and long-term value”. It also states that leaders must “prioritise foundational cybersecurity practices and controls” and “stay actively engaged as threats and guidance evolve”. Engagement is continuous. Stability is conditional.
The UN panel writes that “fragmentation is not fate, and the window to establish shared evidence standards and coordinate global oversight remains open”. It also notes that “open-weight AI models have a sovereignty advantage,” while also acknowledging that “fine-tuning can also degrade or remove safeguards against misuse”. The system is described as simultaneously open, unstable, and unevenly governed.
Widening asymmetry of tech power
Director Ratcliffe states: “We have to move fast. We have to be aggressive.” The Five Eyes statement insists that “the urgency is clear”. The UN panel states that “governing under uncertainty is normal, but AI is distinct”. The language differs. The condition does not.
The irony lies in the widening asymmetry of technological power, which suggests a world in which poorer and structurally lagging states are not simply late adopters but are increasingly governed through systems they neither design nor meaningfully audit, reinforcing a form of dependency in which sovereignty is mediated by external infrastructure.
The periphery does not merely fall behind the centre of power; it becomes a testing ground where opacity, speed, and institutional absence converge into a new form of structural vulnerability.
(The writer is an author based in Colombo)
(The views and opinions expressed in this article are those of the writer and do not necessarily reflect the official position of this publication)