OP-ED: Decoding DeepSeek—Will China’s lean AI model alter global information flows?

Andile Masuku takes stock of the rapid evolution of global AI innovation, exploring its far-reaching implications for journalism, the creative industries, and the legal landscape.

OP-ED: Decoding DeepSeek—Will China’s lean AI model alter global information flows?
Photo by JinHui CHEN / Unsplash

DeepSeek, China's latest entry into the open-source large language model race, has achieved something remarkable: matching Meta's Llama with just one-tenth of the resources, cost and parameters. Questions linger about their methods, though—including possible training on other LLMs without ethical appropriation.

Nevertheless, their architectural breakthroughs, from multi-token prediction to mixture of agents architecture, hint at an intriguing future where efficient, decentralised AI development might challenge the dominance of resource-heavy Western models. It's a development that both excites and unnerves me, particularly when considering its implications for global media and information flows.

As someone who straddles the worlds of journalism, media entrepreneurship and community stewardship in Africa's tech ecosystem, I find myself caught in a peculiar tension. The tools reshaping how we create and distribute information are evolving faster than our ability to understand their implications, while the legal frameworks meant to govern them fall further and further behind the pace of innovation.

This recently came into sharp focus when I encountered a familiar voice in an unexpected context. Listening to a Shirtloads of Science podcast, hosted by the eccentric, science-loving Australian author and media host Dr Karl, I was struck by the insights of fellow Ozzie Natasha Blycha—my English teacher from high school in Zimbabwe in the late 1990s. While she once guided us through literature and essay writing, she had since built an impressive career in international law and technology.

Now CEO of Nooriam (a digital assets infrastructure startup) and managing director of law firm Sterling & Rose, Blycha brings decades of legal experience to the conversation. Her journey through roles as a court associate across multiple countries led to pioneering work in emerging technology law, particularly AI governance frameworks and smart legal contracts. Leveraging this background, she articulated what she terms “the responsible machine problem” with helpful clarity.

The entire legal system, Blycha explains, has been built around human bodies and feelings—entities that can be rehabilitated, punished, or shaped through conventional legal mechanisms. But what happens when AI systems, which can't be "buffeted by feelings or a body", become major stakeholders in our economy?

This isn't sci-fi theory. We're already seeing cases where AI systems are legally required over human alternatives, such as in legal discovery processes. The kicker for me is Blycha’s revelation that some Western courts have determined that technology-assisted review is more accurate and efficient than teams of junior lawyers sifting through documents. In that instance, they've effectively deemed it in the public interest to favour machines over human-based solutions. It’s a sobering harbinger of things to come.

Newsroom revolution

Meanwhile, as Bloomberg News editor-in-chief John Micklethwait observes, AI is getting "under the hood" of journalism itself. His perspective carries particular weight. Having moved from The Economist—one of the Western world's earliest newspapers, founded in 1843—to Bloomberg, he's testified that automation now touches more than a third of their 5,000 daily stories. His "paranoid optimism" resonates with my own experience navigating this rapidly evolving landscape.

As AI promises to revolutionise how we process and deliver information, it raises profound questions about the future of journalism. Honestly, I'm both excited by the possibilities and unnerved by their implications.

African interests

Rewind to October 2022—a lifetime ago in AI terms. An African Tech Roundup podcast episode hosted by Kate Byrne captured fascinating insights from two AI practice specialists at Deloitte. Jania Okwechime, who leads AI and data consulting across Africa, brought technical expertise and regional understanding to the conversation. And, Wessel Oosthuizen, an associate director and AI team lead, offered a tactical perspective.



Back then, they posited that AI could be a powerful equaliser for the continent, potentially allowing us to leapfrog traditional development stages. Their optimism wasn't unfounded, but as someone deeply embedded in Africa's tech ecosystem, I've become more circumspect. Our challenges demand solutions that are applicable to local contexts and constraints. The real question isn't whether AI can solve problems, but whether it can do so reliably with relevance and without perpetuating problematic legacy power structures.

Some might say that this tension has been thrown into sharp relief by recent developments. The appointment of Nigerian billionaire Adebayo Ogunlesi to OpenAI's board has triggered a wave of largely celebratory feedback across African social media. There's a palpable sense that having 'one of us' in the proverbial AI penthouse suite represents more than just a ceremonial win for Africa. Many see it as a glimmer of hope that the continent won't be left behind in the AI revolution that's well and truly underway.

What I see is a deepening consolidation of the profit-driven capital intent currently steering most of Western AI development. Those familiar with private equity (PE) know that when a PE heavyweight enters the chat, it's about to get next-level 'show me the money.' Nothing in Ogunlesi's formidable profile—from his executive tenure at Credit Suisse to founding Global Infrastructure Partners (GIP) and selling it to BlackRock for approximately $12 billion last year—suggests anything other than that he's being brought on to help cement OpenAI's pathway to complete and utter market dominance through commercial exploitation (in the most flattering sense of that term).

This isn't to diminish Ogunlesi's achievements or potential impact, commercial or otherwise. His track record in infrastructure investment could indeed prove valuable for building the foundational systems needed for AI development. Many are waiting to see whether powerful Africans in positions like this can meaningfully influence African participation in the global AI dispensation. But perhaps that's an unfair and incredibly impossible burden for any one person to carry—Africa-interested captain of industry or not. After all, we're talking about accounting for a socioeconomically complex continent, teeming with 1.5 billion people busy building futures here whether or not AI does or doesn't factor into global boardroom calculations. Still, Mr Ogunlesi, sir, we’re all watching.

Creative quandaries

While Africa grapples with its place in the global AI landscape, equally perplexing questions of trust and responsibility are emerging in the creative industries. Tensions between innovation and accountability in AI governance are mounting steadily into the realm of creative work, where the boundaries between human and machine-generated content are becoming increasingly blurred. 

For instance, behold the creative smarts of Daniel Barak, VP and Global Executive Creative Director at New York advertising agency R/GA. As one of the privileged elite granted exclusive trial access to Google's Veo 2, using text and image prompts he created a compelling short film in just 48 hours (spread across 4-5 days). The achievement (among many other similar instances) raises fundamental questions about creativity, ownership and authenticity.

Who actually holds the copyright to AI-generated work like Barak's stunning mini-movie? Is it Barak? Google? The countless original content creators whose work helped train Veo 2? All of the above? Having relied on earning and maintaining public trust to make a livelihood over the years, I approach these uncharted waters with particular caution.

These questions become even more pointed when considering Semrush's latest findings about AI's impact on content marketing and SEO. The results appear promising—68% of businesses surveyed by the digital marketing software company report increased ROI from AI-driven content. Yet, we must ask whether this rush to efficiency comes at the cost of genuine insight and consumer welfare.

Peering into the haze

The convergence of these developments presents a complex, hazy picture. Legal frameworks strain to adapt. Newsrooms transform. Marketers scramble to leverage. African markets explore alternative development paths. And China's resource-constrained AI innovation gains challenge Western assumptions.

The rise of models like DeepSeek hints at a (possible, but unlikely) future where AI development isn't limited to those with the biggest data centres and deepest pockets. China's achievements despite American chip restrictions offer an instructive parallel for African innovators. If resource constraints can drive more efficient approaches to AI development in China, what possibilities might emerge in African markets, where necessity has long mothered invention? This democratisation of AI technology could reshape emerging markets—though questions of oversight and responsible innovation loom large.

The dizzying pace of AI development shows no signs of slowing. Our challenge is to keep up while keeping our heads and our principles firmly in place. As my former teacher Natasha Blycha and journalists the world over might well appreciate, this is a time of many new chapters in the story of human progress—written in thousands of languages—unfolding across an ever-expanding digital universe. And the stakes have never been higher.

Editorial Note: A version of this opinion editorial was first published by Business Report on 16 January 2025.