Sam Altman is reportedly in talks with President Trump to hand the US government a 5% stake in OpenAI, to be distributed to Americans as a kind of AI dividend. It sounds generous until you run the numbers, and the numbers tell a much smaller story than the headline does.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Major Highlights

• The Financial Times reported Altman is discussing giving the US government a 5% stake in OpenAI, an idea he has floated in various forms since 2021.
• OpenAI was valued at $852 billion after its March 2026 funding round, making a 5% stake worth roughly $42.6 billion.
• Split evenly across America’s roughly 133 million households, that stake works out to about $320 per household, hence the story’s title.
• OpenAI is reportedly delaying its IPO until it can hit a $1 trillion valuation, despite still not being profitable.
• The proposal echoes the Alaska Permanent Fund, which shares oil revenue with residents, though Altman rejects the underlying premise that AI wealth, unlike oil, will ever run out.
• Senator Bernie Sanders has floated a far more aggressive version: a 50% government stake in top AI companies.
• The analysis argues the plan may double as political cover, staying in the US government’s good graces matters for AI firms right now, given recent friction between the administration and companies like Anthropic over export and supply chain rules.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

KINI BIG DEAL

Strip away the “AI will make you rich” framing and what’s left is a fairly small number attached to a fairly large amount of political theatre. $320 per household is not nothing, but it is not the wealth-sharing revolution the headline implies either, especially once you remember OpenAI isn’t profitable yet and has been floating this idea in some form for five years without producing an actual mechanism.

The part worth paying attention to, especially for anyone tracking how AI companies navigate government relationships, is the timing. Frontier AI labs are increasingly finding that staying close to government, through equity stakes, favorable trade positioning, or public-facing wealth-sharing gestures, has become part of the cost of doing business in the current political environment, not just in the US but everywhere regulators are deciding how much latitude to give these companies. Nigeria’s own AI Governance Bill delay (covered elsewhere in this batch) is a different flavor of the same underlying tension: governments and AI companies are still negotiating who owes who what, and neither side has fully worked out the terms yet.

Read more: Your family’s $300 stake in OpenAI – MIT Technology Review

Google picked Johannesburg and made its biggest AI announcement for the continent yet. At its first-ever Cloud Summit on African soil, held at the Sandton Convention Centre, the company rolled out five new initiatives in one sitting, everything from subsea cables to a startup lab, and used the moment to tell 3,000 business leaders, developers and policymakers that Africa’s AI economy will be built from the ground up, not bolted on from Silicon Valley.

President Cyril Ramaphosa opened the summit himself, standing shoulder to shoulder with Google executives and Econet founder Strive Masiyiwa. The message made it clear that this is no longer about getting Africa online. It’s about who owns and builds the AI layer on top of that connectivity, and Google wants South Africa to be the launchpad.

Major Highlights

• Google announced Africa’s first Applied AI Lab, based at the Accra AI Community Centre in Ghana, pairing African founders with Google researchers and early access to its latest models. Applications close August 31, 2026.
• Novastar Ventures and other VC partners are backing the Lab, helping identify and fund the African startups that come through it.
• A new Digital Exchange Port in the Eastern Cape will connect Africa directly to Australia via the Umoja subsea cable, plus a new route to India; the first of four such hubs Google has committed to the continent.
• Google partnered with Idris Elba’s Akuna Group on a $1 million AI storytelling initiative for creators across sub-Saharan Africa.
• A R3 million digital innovation centre is being built in Soweto with WeThinkCode, and a new Google for Startups Accelerator cohort for South African founders opens applications on July 21, 2026.
• Google says it has already surpassed its $1 billion commitment to Africa’s digital transformation, ahead of schedule, and its Johannesburg Cloud Region is projected to add $90.6 billion to South Africa’s economy by 2030.

KINI BIG DEAL

This is Google doing what Google does best: showing up with the receipts and the roadmap in the same breath. Five initiatives in one announcement is a statement that Africa has moved from a market to watch to the market to build in. And the choice of South Africa as the base camp is not an accident. Reliable power, a mature cloud region, and an enterprise base already running AI in production (Vodacom, Discovery, Naspers) make Johannesburg the easiest place to stand up infrastructure that then reaches outward to Accra, Nairobi, and Lagos.

But a tension worth sitting with is that the Lab itself is in Ghana, the money and infrastructure conversation is in South Africa, and the loudest AI startup energy on the continent is arguably still in Nigeria. That’s not a contradiction, it’s actually how continental infrastructure gets built, hub and spoke, not everything in one country. The real test isn’t the announcement, it’s August 31, when applications close and we see who Google and Novastar actually pick to build with. If the founders selected skew heavily Anglophone-East-and-Southern-Africa, that tells its own story about where Africa’s AI economy currently means in practice versus in press release.

For African founders, the practical takeaway is simple: apply. Early access to frontier models plus a direct line to Google researchers is the kind of resource gap that has historically separated who scales and who stalls. Sharp people no dey wait for am to blow before dem show face.

You’re scrolling Instagram. A brand you’ve never heard of has a product video that looks like it cost ₦500,000 to produce. The lighting is clean and the motion is smooth. There is even a voiceover. You check the account; it is only three months old, yet it has 8,000 followers. They did not hire a production crew. They used one of the tools on this list.

1. Google Veo 3.1

Google’s flagship video model is currently leading the pack on almost every quality metric. Veo 3.1 generates native 4K video with synchronised audio, meaning the background sounds, ambient noise, and music are generated alongside the video, not added separately afterwards. Prompt adherence is excellent, character consistency across shots is strong, and the output feels cinematic. For anyone serious about AI video quality, this is the benchmark everything else gets measured against right now.

Best for: Narrative scenes, establishing shots, high-quality marketing content.

Pricing: Available through Google AI Ultra Plan at #89,000/month.

Check it out:  https://aistudio.google.com

2. Seedance 2.0

ByteDance built this one, and it currently holds the number-one Elo rating on the Artificial Analysis Video Arena for both text-to-video (1,269) and image-to-video (1,351), ahead of every other model on this list. What makes Seedance 2.0 unique is that it generates video and audio together in a single pass, and it can handle multi-shot narratives from a single prompt, maintaining continuity across scene transitions the way a director would. One catch: generation times can run 60–120 seconds per clip, and content filters are strict following legal pressure from major Hollywood studios.

Best for: Storytellers, multi-scene narratives, creators who prioritise benchmark-proven quality.

Pricing: Available via seed.bytedance.com.

Check it out: https://seed.bytedance.com/en/seedance2_0

3. Kling AI 3.0

Kling is where the conversation about realistic human faces in AI video usually lands. Version 3.0 supports longer clips, multi-shot storyboard mode, and 4K output. The motion of human figures — how people walk, gesture, and move — is noticeably more convincing here than in most competitors. For brands wanting to show real-looking people in their content, Kling 3.0 is the strongest choice at its price point.

Best for: People-centered content, fashion, lifestyle brands, product demonstrations with human models.

Pricing: Approx. #10,000 – #15,000/month — the best value-to-quality ratio in the top tier.

Check it out: https://klingai.com

4. Runway Gen-4.5

Runway is the filmmaker’s tool. While other generators focus on improving output quality, Runway focuses on giving creators more control: Motion Brush lets you direct specific elements of the video independently, camera control is precise, and the integration with professional editing workflows is the most developed in the industry. If you already know what you want a scene to look like and need AI to execute your vision rather than guess at it, Runway is where you go.

Best for: Professional editors, filmmakers, motion designers, advertising agencies

Pricing: Approx. #17,000(Standard) — pro plans available for higher generation volumes

Check it out: https://runwayml.com

5. Hailuo AI 2.3 (MiniMax)

Hailuo generates 1080p video clips in 30–90 seconds — the fastest in the field. At approximately $0.30 per 10-second clip, it dramatically undercuts Runway ($1.20 per clip) and Sora’s former pricing. It is ranked number one for physics simulation on WorldModelBench, meaning objects, liquids, and environmental elements move and interact more realistically than on most competing platforms. The free tier gives you daily credits; enough for two to three videos per day.

Best for: High-volume creators, social media content, budget-conscious professionals, anyone testing AI video for the first time.

Pricing: Free tier available daily; paid plans from approx. #10,000/month. Roughly $0.30 per 10-second 1080p clip.

Check it out: https://hailuoai.video

6. Luma Dream Machine

Luma AI’s Dream Machine consistently produces video with a cinematic quality that distinguishes it from tools optimised purely for realism or benchmark performance. Dynamic perspective shifts, natural lighting changes, and a distinctive visual style make Luma outputs recognisable. It accepts both text and image prompts, so you can feed it a reference photo and let it animate the scene around it. Less about precise control, more about beautiful output.

Best for: Creative directors, artists, brand storytelling, video content that needs to look stunning rather than just realistic.

Pricing: Free tier with limited generations; paid plans from approx. #42,000/month.

Check it out: https://lumalabs.ai/dream-machine

7. Pika

Pika built its name on lip-sync, specifically Pikaformance, a tool that takes a still image and syncs it convincingly to audio or dialogue. For talking-head content, product explainers, or social media clips where you need a character to speak directly to camera, Pika is the specialist. It is less suited for complex narrative video, but for its specific use case it is ahead of the pack.

Best for: Social media content, talking-head videos, short-form clips, anyone who wants a character to speak in a video.

Pricing: Free tier available; paid plans from approx. #11,000/month.

Check it out: https://pika.art

8. Synthesia

Synthesia is the market leader for a specific and valuable use case: professional videos using realistic AI avatars as presenters. It is widely used for corporate training, onboarding videos, e-learning content, and internal communications; anywhere you need a professional-looking presenter without a camera crew. Over 160 languages are supported, which makes it especially relevant for African businesses producing multilingual content. The avatars are not for creative filmmaking, but for professional presentation they are convincing.

Best for: Businesses, HR teams, e-learning developers, educators producing multilingual instructional content.

Pricing: From approx. #25,000/month (Starter) — enterprise plans available.

Check it out here: https://www.synthesia.io

9. HeyGen

HeyGen occupies the space between Synthesia’s corporate rigour and the creative flexibility of the text-to-video tools. It is built for marketers: create an AI avatar of yourself or choose from a library of realistic avatars, add a script, and generate a professional video in minutes. The platform is widely used for product launches, explainer videos, and personalised video outreach. It also has strong translation features — you can record a video in English and have HeyGen re-lip-sync it in French, Swahili, or Yoruba.

Best for: Marketers, founders, educators, content creators who want to appear on camera without filming.

Pricing: Free plan available; paid plans from approx. #40,000/month.

Check it out: https://www.heygen.com

10. InVideo AI

InVideo AI is the most accessible tool on this list. You describe the video you want in plain language, choose a format (YouTube, Instagram Reel, explainer, ad), and InVideo AI writes the script, selects stock footage or AI-generated visuals, adds voiceover, and produces a ready-to-publish video. It is not the most powerful creative tool, but for a small business owner, a teacher, or a creator who has never edited a video and needs something professional today, it is the most practical starting point. It also has an active African creator community.

Best for: First-time video creators, small business owners, social media managers, educators.

Pricing: Free plan available with watermark; paid plans from $20/month.

Check it out: https://invideo.io

KINI BIG DEAL

That brand you were watching on Instagram probably spent less than #20k on that video, maybe nothing at all.

Four of the tools on this list — Hailuo AI, Pika, HeyGen, and InVideo — have free tiers that are genuinely usable. Not the kind where you click the button and discover you need to upgrade before anything actually happens. Hailuo gives you daily credits. Pika’s free plan produces real clips. HeyGen lets you create a video before you enter a card number. InVideo AI’s free plan produces content you can actually publish. If you are in Africa and have not started experimenting with at least one of these tools, the barrier is not money. It is just inertia.

The other thing worth knowing — and this is specifically for African businesses — two tools on this list can take a video you record in English and re-lip-sync it in Yoruba, Hausa, French, or Swahili. Synthesia supports over 160 languages. HeyGen does the same. You do not re-record. You do not hire someone to translate and reshoot. You upload, click, and wait a few minutes. For any business, school, or creator that needs to reach an audience in a language they do not present in fluently, that capability is live right now.

The account you were scrolling past already knew all of this. Now you do too. Run!

Walk through any market in Lagos or Nairobi and you will find the same thing: a trader holding a phone, managing orders on WhatsApp, posting products on Instagram, trying to keep track of what sold and what is still in stock, all in their head, across three different platforms, with no data, no analytics, and no way to know if their business is actually growing.

That is how most of Africa’s eCommerce works. The continent’s online shopping market is worth over $55 billion and is projected to reach $112 billion by 2029. But most of the people driving that market are doing it from their phones, without a proper digital storefront, because the tools that exist were not built with them in mind.

Payaza, a Pan-African fintech company, launched Shopaza in Lagos on June 18 to change that. The idea is simple: give African merchants and diaspora sellers a real digital store, built on a payment system that already works, with AI doing the heavy lifting on setup.

Major Highlights

• Payaza launched Shopaza, an AI-powered eCommerce platform for small businesses, on June 18, 2026 in Lagos.
• Shopaza is live across 23 countries — in Africa, North America, and Europe — targeting both African merchants and diaspora sellers.
• The platform runs on Payaza’s own payment infrastructure, allowing instant settlement — a major advantage over platforms that depend on third-party payment processors and slower payout timelines.
• The headline feature: an AI store builder — merchants drop a product image, and the AI automatically sets a competitive price, populates product descriptions, and adds size, colour, and variation details, requiring no manual entry.
• Stores on Shopaza are fully customisable, with multiple templates for physical products, digital products, and bookings — unlike most eCommerce platforms that give every store the same look.
• The core problem Shopaza addresses: merchants selling on social media have no data on who bought what, no inventory management, and no way to see which products are performing: just messages, bank transfers, and guesswork.

KINI BIG DEAL

The social media storefront is one of the most creative workarounds the African entrepreneurial spirit ever invented. Africans turned Instagram DMs and WhatsApp broadcast lists into legitimate sales channels before most of the world’s eCommerce platforms even knew those markets existed. The ingenuity is real.

But ingenuity and infrastructure are two different things. A business that cannot see its own data cannot grow beyond what one person can hold in their head. You cannot restock intelligently if you do not know what sold. You cannot plan marketing if you cannot see when your customers are active. You cannot expand if your entire business lives in three apps on a phone that could get stolen tomorrow.

What Shopaza is offering — especially the instant settlement piece — is not just convenience. It is cash flow. In Nigerian business, the gap between “the customer paid” and “the money is in my account” can mean you cannot restock, cannot pay your supplier, cannot keep the operation running. Payaza’s in-house payment infrastructure closing that gap is the kind of practical detail that makes the difference between a merchant who scales and one who stays stuck.

The AI store builder is also worth pausing on. For a first-time digital merchant, setting up a proper eCommerce store has always meant navigating templates, writing product descriptions, researching pricing, and uploading images — hours of work that a trader with twenty products and a market stall to run simply does not have. Dropping a photo and letting the AI handle the rest is not a gimmick. For the right merchant, it is the difference between signing up and giving up.

Read more: Techpoint Africa — Payaza launches Shopaza, integrates AI to simplify eCommerce

Until next time, stay curious.

Anthropic Launched Its Most Powerful AI Ever. The US Government Pulled It Three Days Later.

On June 9, 2026, Anthropic launched Claude Fable 5;  the most capable AI model the company has ever released. It sat above the Opus class in a new tier called Mythos. It had a one-million-token context window, 128,000 output tokens, and always-on adaptive thinking: a mode that makes the model automatically double-check its own reasoning before responding, every single time.

It was available for exactly seventy-two hours.

On June 12, the US government issued an export control directive ordering Anthropic to suspend all access to Fable 5 and its companion model, Mythos 5. The stated reason was the discovery of a potential jailbreak method. Anthropic complied and the models went dark. It is the first time a government has ever used export controls to pull a commercial AI model from the market.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Major Highlights

• Claude Fable 5 and Mythos 5 : Anthropic’s first Mythos-class models, representing a new tier above Opus, launched on June 9, 2026, available on the Claude API, Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry the same day.
• The models featured a 1-million-token context window, 128K output tokens, and always-on adaptive thinking, which basically means a reasoning mode that cannot be disabled, built to make the model catch its own errors in real time.
• Safety classifiers were included that automatically fall back to Claude Opus 4.8 for high-sensitivity domains such as cybersecurity and biology.
• On June 12, three days after launch, the US government issued an export control directive requiring Anthropic to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including Anthropic’s own non-US employees.
• Because enforcing the directive only for foreign nationals was practically impossible without disabling the models entirely, Anthropic suspended access for all users globally.
• The government’s concern: A method of bypassing Fable 5’s safety guardrails; a jailbreak had been discovered and demonstrated to officials.
• Anthropic publicly disagreed with the decision, stating their review of the jailbreak demonstration found it exposed only a small number of previously known, minor vulnerabilities, and that pulling a model used by hundreds of millions of people was a disproportionate response.
• Anthropic says it is working to restore access and is engaged with the government to resolve the situation.
• All other Anthropic models — Claude Opus 4.8, Sonnet, Haiku — remain fully available and unaffected
• This is the first-ever government-ordered suspension of a commercial AI model using export control authority.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

KINI BIG DEAL

What the US government did was a safety move. A jailbreak was demonstrated on the world’s most powerful AI model; one with a one-million-token context window and always-on reasoning. Pulling it first and asking questions later is arguably the correct instinct. You do not leave a compromised system running while you investigate how bad the breach is.

The problem is not that they acted. The problem is what acting looks like from everywhere else.

From Lagos, this is what it looks like: the model was available on Monday. By Thursday, it was gone without a warning. No process for governments, universities, hospitals, or businesses outside the United States to be part of the conversation, or even to receive advance notice. A safety decision made in Washington became an immediate operational reality for hundreds of millions of people who had no input and no recourse.

If a jailbreak is serious enough to pull a frontier AI model, then the safety of these tools is a global concern. But the governance of that safety is not global. It happens in one country, under one regulatory framework, with one set of interests at the centre. The rest of the world participates as users, not as stakeholders.

That is what this week made visible. Fable 5 will likely return, but the structure that made this possible — where a safety call in Washington is automatically a shutdown for everyone else — that does not go away when it does.

The Africa Deep Tech Challenge is back, and this year’s edition has a very specific target.

Every year, the Africa Deep Tech Foundation runs one of the continent’s most credible open innovation competitions. The 2025 edition drew over 2,800 innovators, produced 400+ project submissions, and awarded $20,000 in equity-free grants to builders solving real African problems with genuine technical depth. The 2026 edition is now open, and the theme is one that every African developer has felt personally: building AI that does not require infrastructure most of the continent does not have.

The Laptop LLM Challenge 2026 is asking builders to create large language model solutions that run efficiently on consumer-grade laptops. AI that works on the machine already sitting on your desk, without requiring GPU farms or expensive cloud subscriptons, and addresses real-world African problems while it is at it.

Applications are open now and the deadline is Friday, July 24, 2026.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

Major Highlights

• The Africa Deep Tech Foundation: The organisation behind the annual Africa Deep Tech Challenge has launched the Laptop LLM Challenge 2026, calling on innovators, researchers, developers, and AI builders across the continent and diaspora to compete for a $20,000 prize pool in equity-free grants.
• The challenge has one core requirement: build an LLM solution that runs efficiently on consumer-grade laptops — hardware that ordinary African professionals actually own — without dependence on expensive cloud infrastructure.
• Submissions must address real-world African problems, meaning the technical challenge and the social relevance are baked into the same brief.
• The challenge is open to individuals and teams: students, researchers, startup founders, and independent builders all qualify.
• Winners receive not just cash, but mentorship, investor access, and post-challenge support, as was the case in the 2025 edition.
• The Africa Deep Tech Challenge has real precedent: the 2025 edition drew over 2,800 innovators and more than 400 unique project submissions, with live finals producing winners including FarmSpeak (smart tools for farmers), Cure Bionics (bionic prosthetics), and Muscle (digitising African small retail).
• Applications close Friday, July 24, 2026 at africadeeptech.org/challenge-2026.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━

KINI BIG DEAL

Hardware problem is not talked about enough because it is both a money and a structural problem. 

Most of the world’s frontier AI models were built by organisations with access to tens of thousands of GPU chips, data centres the size of warehouses, and electricity infrastructure that most African countries have not yet stabilised. The assumption baked into those systems — that powerful AI requires powerful hardware — is not a law of physics. It is a design choice made by people who have access to resources most of the world does not have.

The Laptop LLM Challenge is asking a different question: what if we designed AI specifically for the hardware constraints that African builders actually face? What if resource efficiency was not a compromise but the brief itself?

This is not a small reframe. The 2025 edition of the Africa Deep Tech Challenge was built around exactly this theme and it produced genuinely innovative work. FarmSpeak, the grand prize winner, built tools to help farmers using the kind of hardware farmers actually have. The pattern is consistent: when you design for the constraint, you often build something more useful for the continent than when you import a solution that was designed for someone else’s context.

Twenty thousand dollars is a life-changing sum, but the recognition, the investor access, and the credibility of winning a challenge explicitly designed for African conditions — those are the actual prizes.

If you are a developer, researcher, student, or builder anywhere on this continent and you have been waiting for a reason to build something serious with AI: abeg, this is your sign. The deadline is July 24. That is five weeks from today.

Apply here: africadeeptech.org/challenge-2026 

Read more: TechBuild Africa | Africa Deep Tech Foundation