By Sikhosonke Mayekiso

Imagine this: You are scrolling through your feed on a lazy Sunday in Johannesburg, and there it is, a video of President Cyril Ramaphosa, his voice steady and authoritative, announcing a radical new policy that slashes jobs in the mining sector. Outrage erupts. Hashtags trend, protests brew, and the rand takes a nosedive. But wait! it’s all fake. A deepfake, conjured by AI with a few clicks, anonymous and untouchable. In South Africa, where our hard-won democracy still echoes the scars of apartheid’s censorship, AI promises “unlimited” freedom of expression. Yet, as we’re seeing in real-time scams, election meddling, and reputational assassinations, this digital sorcery is less a liberator and more a Trojan horse for deception. Let us unpack this critically: Is AI’s veil of anonymity empowering voices, or eroding the very foundations of truth in our rainbow nation?

First, the allure. AI tools like deepfakes democratise expression in ways our Constitution’s Bill of Rights could only dream of. Section 16 guarantees freedom of speech, but historically, that’s been a privilege for the powerful. Think state-controlled media under apartheid or today’s gatekept broadcasters. Now, anyone with a smartphone and free software can “speak” through fabricated videos, audio, or images. Behind the screen, anonymity reigns supreme. A whistleblower in Soweto could expose corruption by morphing into a fictional character, shielding themselves from retaliation. Satirists could lampoon politicians without fear, amplifying marginalised voices in a country where economic inequality silences millions. It’s the ultimate equaliser: No printing press needed, no broadcaster’s approval. In a nation grappling with youth unemployment and digital divides, AI hands the mic to the masses, encouraging creativity and critique that feels boundless.

But here’s the critical rub. This “unlimited” freedom is a illusion, one that’s already fracturing our society. Deepfakes don’t just express; they deceive, and in South Africa, deception hits hard. Take the 2024 case of Leanne Manas, the beloved SABC anchor whose face was hijacked in deepfake ads peddling bogus weight-loss scams on Facebook and TikTok. Or Elon Musk, our SA-born tycoon, starring in a fabricated video luring locals into fraudulent investments promising sky-high returns. These aren’t harmless pranks; they are cybercrimes that exploit trust in a country where financial desperation runs deep. Victims lose savings, reputations shatter, and the anonymous perpetrators vanish into the ether. Worse, enforcement lags woefully behind. South Africa’s laws, like the Cybercrimes Act of 2020, which criminalises malicious deepfakes, or the Films and Publications Amendment Act, targeting fake news distribution – look solid on paper. But as legal experts point out, the real bottleneck is implementation: Overburdened police, tech-illiterate courts, and platforms like Meta that drag their feet on takedowns. Justice delayed is justice denied, especially when deepfakes spread faster than wildfire in the veld.

Now, zoom in on politics, where the stakes skyrocket. Our 2024 elections were a testing ground for AI mischief. Remember the Umkhonto we Sizwe party’s deepfake of Donald Trump “endorsing” Jacob Zuma? It went viral among 26 million social media users, sowing confusion and potentially swaying votes in a fractured political landscape. In a nation still healing from state capture and ethnic tensions, such manipulations aren’t just “expression” – they’re weapons. They undermine Section 19’s right to free, fair elections, eroding trust in institutions already battered by corruption scandals. Critics argue AI amplifies echo chambers, where anonymous bots and deepfakes reinforce biases, from xenophobia to racial divides. And let’s not sugarcoat it: In Africa’s digital frontier, with limited AI governance, we’re ripe for foreign interference. Think Russian bots in the US elections, but tailored to SA’s fault lines with fake videos inciting service delivery riots or stoking farm murder myths. Freedom of speech? Sure, but at what cost to our fragile unity?

The anonymity angle is particularly insidious. AI lets creators hide “behind” algorithms, evading accountability that traditional media demands. In SA, where social media is a lifeline for expression amid poverty, this veil empowers trolls and scammers alike. Yet, it clashes with our constitutional limits: Speech isn’t absolute; hate speech, defamation, and incitement are outlawed. Deepfakes blur these lines, turning privacy invasions into “art.” Nonconsensual deepfake porn, for instance, ravages women’s rights, yet our performers’ protection laws are being stretched thin to combat it. Data protection laws, like POPIA, offer some shield, but they are patchwork. We need a rethink, particular proactive regulation, like the EU’s AI Act, adapted to SA’s context, with mandatory watermarking and AI literacy in schools.

Don’t get me wrong – banning AI outright would echo apartheid’s censorship, stifling innovation in a country hungry for tech jobs. But unchecked, deepfakes risk a “post-truth” SA, where every video is suspect, and real injustices (like farm attacks or police brutality) get dismissed as fakes. The Legal Resources Centre’s push for ECTA takedown notices is a start, and is notably an empowering move.

Disclaimer: This post reflects my personal views only and is not legal advice. Consult a qualified professional for your specific legal and related queries.