If you've ever used ChatGPT, Gemini, or Siri and wondered who decided it was going to work this particular way — mostly, it was the company that makes it. OpenAI for ChatGPT. Google for Gemini. Apple for Siri. Engineers in San Francisco and Seattle making calls about what a few billion people experience.

That's been changing, slowly. This week it changed a little more.

The UN and its International Telecommunication Union launched the AI for Good Global Commission on July 2 — a formal body bringing together more than 40 world leaders, government officials, and tech executives to steer how AI gets used globally. Co-chairs: Salesforce CEO Marc Benioff and Rwandan President Paul Kagame. First meeting: July 8 in Geneva.

The headline problem they're organized around: 2.2 billion people worldwide are still offline. That's one in four people on Earth. And AI — the thing that's supposed to help with healthcare, education, disaster response, food security — requires internet access to use.

What "voluntary alignment" actually means

Here's the honest version of how this works, because "commission" sounds powerful and "voluntary alignment" is the fine print.

The commission isn't a regulator. It can't pass laws. It can't fine companies. It cannot tell OpenAI to deploy its models in rural Kenya, or force Google to make Gemini available in a language spoken by 20 million people. What it can do is:

• Publish frameworks and guidelines that governments can choose to adopt
• Use the combined status of its members to put pressure on companies
• Convene the people who do have real authority in the same room at the same time

Think of it like a PTA for global technology — the people with authority over the building (governments) sit next to the people who built the building (the tech executives), argue about priorities, and some of it results in policy change and some of it results in a report nobody reads.

What makes this particular body different from the previous 50 AI governance attempts is that the tech CEOs are literally on the commission, not just being lobbied by it. Founding members include Jensen Huang (Nvidia), Andy Jassy (Amazon), Brad Smith (Microsoft), Jack Clark (Anthropic co-founder), and Aidan Gomez (Cohere CEO). When the commission produces a framework, the people who make the products it covers are in the room when it's written.

That's an argument for the commission and against it, at the same time.

Source spread

• Salesforce — AI for Good Global Commission press release — [hype]. Official launch materials. Optimistic framing, but solid on membership list and mission statement.
• Axios — Exclusive: UN launches AI for Good commission — [builder]. Benioff quotes and the behind-the-scenes account of how the commission was convened.
• ITU — Official press release — [hype]. Primary UN source; official framing of the 2.2 billion figure and the institutional precedents.
• Common Dreams — UN Creates AI for Good Commission full of big tech execs — [skeptic]. Worth reading as a counterweight. Makes the structural conflict argument clearly.

What's actually fine / What deserves a side-eye

What's actually fine:

Someone finally got the tech companies and world governments in the same room at a level of seniority where things can actually happen. The previous versions of this coordination were either government-only (companies ignored them) or company-only (governments ignored them). This is both.

Paul Kagame's co-chair role isn't decorative. Rwanda has been one of the most aggressive African governments at investing in digital infrastructure, and Kagame has a real track record of implementing programs, not just announcing them. His presence brings a head of government who knows what it takes to build connectivity in places that don't have it.

The 2.2 billion figure anchors the commission to a concrete problem. Governance bodies that talk about AI "in general" tend to produce frameworks useful for no one in particular. The offline population number is a measurable thing that either gets better or doesn't.

What deserves a side-eye:

Benioff's launch quote — "AI is the most profound technological transition in history. And our values have to guide every step, because responsibility is the core of AI ethics" — sounds right and means nothing until there's a specific decision attached to it. Whose values? Enforced how? With what consequences for non-compliance?

The mode is voluntary alignment. Every company participates as much as it finds convenient. The history of voluntary AI governance, from the Montreal Declaration to the G7 Hiroshima AI Process to the 2023 safety commitments, is that companies sign on enthusiastically and the commitments quietly expire when no enforcement follows.

African delegates at the Geneva summit have already raised objections that the founding membership doesn't adequately represent the regions that will be most affected by AI access gaps. The 2.2 billion offline are disproportionately in Africa and South Asia. The CEOs at the table are disproportionately from Silicon Valley. That's a structural tension the commission has to navigate before it can claim to speak for the people it's supposedly helping.

What to do about it

Five things that are actually useful to do with this news:

• Distinguish frameworks from commitments. When the Geneva summit produces outcome documents, look for whether companies made specific commitments with dates, or signed onto a principles document. The first type has accountability built in. The second type is wallpaper.
• Follow the African pushback. The most substantive political story inside this commission is whether the regions most affected by AI access gaps get meaningful representation in how the frameworks are written. African delegates have already raised this publicly. Watch whether it changes the outcome documents or gets smoothed over.
• The 2.2 billion number is worth sitting with. If AI is going to matter for healthcare, education, and economic opportunity globally — and it will — the tools have to reach the people who need them most, not just people who can afford $20 a month and have reliable internet. Whether the commission is addressing the economics of that or just the governance layer is a real distinction.
• The free tools already exist. Whatever the commission produces, practical AI tools that work in low-bandwidth environments — Khan Academy's AI tutor, open-source health screening models, Meta's Llama-based tools — are live right now, without needing any Geneva summit to do anything. The governance story is slow. The tools are fast.
• Ask your AI about this. It's not circular to ask ChatGPT or Claude to explain why AI access is unequal globally and what this commission is actually trying to do about it. These tools are reasonably good at explaining policy topics in plain language. That's a good use of them.

Further reading

• ITU — AI for Good Global Commission official announcement — primary UN source
• Salesforce — founding member press release with membership list
• Axios — Exclusive: UN launches AI for Good commission — Benioff quotes and inside account
• Tech Financials — Africa seeks stronger AI influence — the pushback worth reading
• Common Dreams — AI for Good commission full of big tech execs — skeptical read on structural conflicts