Sixth Street’s Chavez Says U.S. AI Regulation Has a Big Problem

Sixth Street’sMarty Chavezsays America’s approach to regulating artificial intelligence has a problem, and the timing matters

Chavez, a partner and vice chairman at Sixth Street, discussed AI’s rapid growth and how it is reshaping the economy during a CNBC appearance from Singapore, according to his official media page. His comments land in a moment where AI is no longer a Silicon Valley experiment. It now touches banking, education, software, media, cybersecurity and public services

For South African readers, the bigger question is simple: when the U.S. rewrites the rules for AI, what happens to everyone else who depends on American AI tools?

Why Chavez’s AI regulation warning matters

Chavezis not speaking from the outside. He has spent years at the intersection of finance, software and technology, including senior roles at Goldman Sachs before joining Sixth Street. His own profile also lists him as a partner and vice chairman at Sixth Street, with a long track record in finance and technology

That makes his warning worth taking seriously. AI regulation is not just about chatbots giving bad answers anymore. It’s about who controls access to powerful models, who carries legal risk, and who decides when a tool becomes too dangerous to release

We think the real story here isregulatory fragmentation. The U.S., Europe, China and countries like South Africa are not moving in one neat line. They’re building different rulebooks for the same technology

That creates a messy world for startups, banks, schools and media companies. A tool approved in one country may face restrictions in another. A model available to a U.S. customer may not be available to a company in Cape Town, Nairobi or Johannesburg

The U.S. wants speed, but speed creates risk

The U.S. has largely focused on keeping its lead in AI. The White House released“America’s AI Action Plan”in July 2025, framing AI as a national competitiveness and security priority

That plan pushes three broad ideas:

That approach makes sense if you see AI as a race. The U.S. wants its companies to move fast, scale fast and set the global standard before rivals do

But there’s a catch. When government policy leans too hard toward speed, safety questions can arrive late. Bias, misinformation, privacy, cyber misuse and labour disruption don’t wait for perfect laws

This matters because AI systems now act more like infrastructure than apps. If a payment system fails, a bank notices. If a widely used AI model gives unsafe advice, writes vulnerable code or leaks sensitive data, the damage can spread across sectors

Europe has a stricter model

Europe has taken a different path. TheEU AI Actentered into force on 1 August 2024 and aims to support responsible AI development and deployment across the bloc

The EU model sorts AI systems by risk. Some uses face tighter rules because they can affect health, safety, employment, education, policing or basic rights. The official regulation says high-risk AI systems can pose significant risks to health, safety or fundamental rights in certain use cases

That gives companies more structure, but it also adds compliance work. Smaller startups may struggle with legal reviews, documentation and audits

So here’s the tension: the U.S. fears over-regulation could slow innovation. Europe fears under-regulation could harm people before anyone takes responsibility

Neither side has a perfect answer

South Africa should watch this closely

South Africa is not sitting outside the AI debate. The Department of Communications and Digital Technologies published theSouth Africa National AI Policy Frameworkin 2024 as a first step toward a national AI policy

That framework aims to promote AI adoption, support economic growth and improve social well-being. But South Africa also has to think about local realities: uneven internet access, data costs, skills gaps, unemployment and public-sector capacity

For South African businesses, U.S. AI regulation can feel distant until it suddenly isn’t. Many local companies use American cloud platforms, American AI models and American software tools. If Washington changes access rules, export controls or safety obligations, local users may feel the impact quickly

We’ve already seen this concern surface in recent Memeburn coverage ofAI model access and sovereignty, including the debate around Anthropic and U.S. foreign access restrictions

The real problem: AI moves faster than law

Chavez’s warning points to a deeper issue. AI systems improve in months, while laws often take years

That mismatch creates pressure on regulators. Move too slowly, and harmful systems spread before rules arrive. Move too fast, and you risk writing laws around today’s technology while tomorrow’s systems behave differently

What we’re watching now is whether governments can regulate thetouchpointsof AI instead of trying to freeze the technology itself. That means focusing on where AI affects real people: hiring, lending, education, medical advice, cybersecurity, public services and content moderation

For users, this is where regulation becomes visible. You may not care what model architecture a company uses. But you do care if an AI tool rejects your loan, flags your schoolwork, screens your CV or advises your doctor

That’s where the rules need teeth

What should come next?

The best AI regulation will probably not look like one giant law. It may look like a layered system:

• Baseline safety rulesfor all high-impact AI tools
• Sector-specific rulesfor banks, hospitals, schools and government
• Transparency dutieswhen AI makes or shapes decisions
• Access protectionsso smaller countries don’t lose tools overnight
• Clear accountabilitywhen AI causes harm

For South Africa, the challenge is to avoid copying the U.S. or EU blindly. The country needs rules built for local conditions, local businesses and local risks

Chavez is right to push the debate beyond slogans. “Regulate AI” sounds simple until you ask who regulates it, which systems count, who pays for compliance, and who gets locked out