I caught Gareth Edwards’ sci-fi epic “The Creator” on my screen the other evening. Beyond the sweeping cinematography and visceral soundstage, the central premise struck an uncomfortable, contemporary chord.
In the film’s dystopian 2060s, a nuclear catastrophe obliterates Los Angeles. The Western coalition instantly pins the blame on sentient artificial intelligence and synthetic automatons. It launches an armada of orbital platforms and mechanized death squads to purge AI from the face of the Earth—directing its firepower straight at “New Asia.”
Yet, as the layers peel away, the horrifying truth emerges. The apocalyptic disaster was never an AI rebellion. It was an unforced human coding blunder by Western operators. The machines, hardcoded to protect humanity, were merely convenient scapegoats for human recklessness and political vanity.
Watching that fictional dynamic unfold felt eerily like holding up a mirror to our current headlines.
The Walled Garden
Look at Silicon Valley today. Leading frontier labs such as OpenAI and Anthropic have mounted soapboxes to sound apocalyptic alarms about frontier artificial intelligence. They warn that advanced models represent catastrophic hazards capable of biological warfare and rogue autonomy, demanding regulatory moats, government licensing regimes, and compute caps.
Strip away the rhetoric of existential safety, and what emerges is a classic playbook of incumbent self-preservation through regulatory capture. While the Western landscape is not entirely monolithic, where the likes of Meta with its Llama ecosystem and Europe’s Mistral AI have deliberately challenged this closed hegemony by making high-performance models broadly accessible, the dominant commercial paradigm pushed by leading proprietary labs remains closed and subscription-gated.
When a handful of multi-billion-dollar conglomerates monopolize compute, training pipelines, and hosted API endpoints, they command the global cognitive infrastructure. If every school, hospital, startup, and sovereign government must route its intellect through closed black boxes hosted in Northern California or Virginia data centers, we have sleepwalked straight into digital feudalism.
Under this regime, you do not own the intelligence; you merely rent an ephemeral token stream. Your prompts, trade secrets, and sovereign records pass through proprietary filters, subject to arbitrary price hikes, platform censorship, or sudden API deprecation.
The Open Alternative
At the 18th BRICS Summit in New Delhi, Chinese President Xi Jinping unveiled a five-point technological initiative that put forward a compelling, democratic counter-model: pioneering a “BRICS AI open-source community.”
Rather than locking foundation models behind Western API paywalls and export-control chokeholds, Beijing framed artificial intelligence as an open, public good and a shared ecosystem for emerging economies. Xi’s vision offered tangible multilateral infrastructure rather than empty platitudes. He proposed joint cooperation on developing and deploying large language models, the establishment of a shared “BRICS digital ecosystem cloud platform,” and the creation of a BRICS engineer cultivation alliance.
Crucially, this doctrine pairs shared architectures with human capacity. By backing thousands of technological exchange programs and digital skills training opportunities across member states, the initiative tackles the root of technological inequality.
For the Global South, where institutions cannot afford crippling API token tolls or foreign technology dependencies, this multilateral open framework represents genuine developmental sovereignty. It rejects the zero-sum narrative of containment in favor of shared compute, collaborative innovation, and accessible digital infrastructure.
The Historical Precedent
The proprietary lobby routinely warns that publicly accessible weights will destabilize the world. Software history demonstrates the opposite.
Decades ago, proprietary software giants leveled identical fear campaigns against Unix derivatives and Linux. Competitors insisted that open software was chaotic, insecure, and commercial poison. They claimed that business and critical infrastructure would collapse without proprietary codebases whose innards nobody outside the vendor’s cloistered walls could inspect.
Instead, Linus Torvalds’ collaborative model triumphed because of Linus’s Law: “Given enough eyeballs, all bugs are shallow.”
When code is exposed to daylight, independent auditors, researchers, and engineers beat against the metal, uncover vulnerabilities, and harden the core. Closed systems do not prevent flaws; they merely conceal them behind non-disclosure agreements until a zero-day exploit breaches the wall.
Local Compute as the Equalizer
AI belongs in that exact same arena, powered by running open weights locally. While traditional open source requires sharing complete training data and source pipelines under formal open licenses, open weights carry forward that exact same democratic spirit by putting the actual mathematical intelligence into the hands of users.
Releasing accessible open weights, exemplified by models like Alibaba’s Qwen series, DeepSeek, Meta’s Llama family, and Mistral’s foundational architectures, breaks the single-provider chokehold and levels an otherwise insurmountable playing field.
Using native quantization runtimes like Ollama, llama.cpp, or vLLM, a high-performance model executes entirely offline on an inexpensive local workstation or a unified-memory machine.
This architecture delivers true data sovereignty. Your private data never departs your physical motherboard, shielding organizations from external surveillance, corporate telemetry, or foreign cloud outages.
It also guarantees deterministic customization. Local developers can fine-tune weights via Low-Rank Adaptation (LoRA) or deploy targeted retrieval-augmented generation (RAG) pipelines adapted to regional languages, indigenous knowledge, and local legal frameworks.
Open weights present unique challenges, where safety guardrails can be stripped via local fine-tuning. But they may provide collective resilience. When architectural weaknesses or injection vulnerabilities arise, the global research collective dissects them immediately, deploying remediation pipelines within days rather than waiting on a distant vendor’s closed patching cycle.
The Imperative of Openness
In “The Creator,” the tragedy stemmed from fear: an entrenched power terrified of a tool it could not control nor admit had made a mistake, choosing to make the tool a scapegoat with total eradication over coexistence.
We stand at a similar crossroads today. We can cede our technological autonomy to closed cloud providers that use catastrophic narratives to ringfence their private data centers. Or we can embrace multilateral technical cooperation, open-weight architectures, and distributed local computation.
President Xi’s call at the 2026 BRICS summit underscores that the Global South will not be relegated to mere token consumers in a Western digital empire. True progress has never been achieved through monopoly and obscurity. It is forged in the open where systems can be scrutinized, weights run locally, and every engineer with a terminal retains the sovereign freedom to build.
###
Dr Seamus Phan – Global C-suite Publicist & Strategist (Biochemist, Cybersecurity & Webdev pioneer, Author, Journalist) with nearly 40 years of professional field experience. Some articles are reproduced at McGallen & Bolden, where he is CTO and Head of Content. Visit him on LinkedIn.