Who Governs the AI Economy? Regulation, Data, Infrastructure and the New Battle for Technological Control

Author’s Note and Disclaimer:
Zion Zhao Real Estate | 88844623 | 狮家社小赵 | wa.me/6588844623 |  https://linktr.ee/zionzhao

This post is for general information, education, and market literacy only. It does not constitute financial, investment, trading, legal, tax, accounting, or other professional advice, and is not an offer, solicitation, recommendation, or endorsement. Views expressed are personal, general in nature, and subject to change without notice. While reasonable care is taken, no representation or warranty is given as to accuracy, completeness, or reliability. Readers should conduct independent due diligence and seek professional advice. To the fullest extent permitted by law, no liability is accepted for any loss arising from reliance on this material. 

Can the AI Industry Regulate Itself? The Real Battle Is Over Trust, Infrastructure and Control

Artificial intelligence is no longer merely a software sector. It is becoming economic infrastructure.

AI now influences financial transactions, corporate data, scientific research, software development, energy policy and national competitiveness. That is why apparently unrelated developments, including a proposed AI self-regulatory organisation, Stripe’s reported approach for PayPal, Apple’s lawsuit against OpenAI, an AI coding-tool privacy failure, China’s rapid model progress, New York’s data-centre moratorium and an experimental anti-ageing enzyme, all point to the same conclusion:

The next stage of AI will be defined not only by capability, but by governance, trust and control.

AI Cannot Regulate Itself Alone

Google DeepMind chief executive Demis Hassabis has proposed a United States-led AI standards body modelled partly on the Financial Industry Regulatory Authority, or FINRA. Such an organisation could be industry-funded, staffed by technical experts and responsible for evaluating frontier models before deployment.

The proposal has merit because conventional government regulation often moves too slowly for AI. Legislators may struggle to understand rapidly changing model architectures, benchmarks and deployment risks. A rigid approval system could become obsolete before it is implemented.

However, describing the proposal as pure self-regulation is misleading. FINRA does not simply allow financial firms to supervise themselves. It operates under the authority and oversight of the United States Securities and Exchange Commission, with rulemaking, examination and disciplinary powers (FINRA, n.d.).

The appropriate model for AI is therefore co-regulation, not voluntary industry promises.

AI companies possess essential expertise, but they also face commercial incentives to release models quickly, protect proprietary information and shape regulations that favour incumbents. Research on AI governance warns that complex compliance systems may produce regulatory capture by allowing dominant companies to influence rules or impose costs that smaller competitors cannot afford (Wei et al., 2024).

A credible AI regulatory body should therefore include independent technical testing, public-interest governors, representation for start-ups and open-model developers, post-deployment monitoring, incident reporting, enforceable sanctions and government oversight.

The industry must help design the regulator. It cannot be the sole judge of its own conduct.

Stripe’s PayPal Interest Is About Owning the Payment Stack

Stripe and private-equity firm Advent International reportedly offered approximately US$53 billion for PayPal. Contrary to early reporting, Block is not part of the current proposal, although it reportedly participated in an earlier approach. PayPal has not accepted the offer, and its board reportedly considers the valuation inadequate (Reuters, 2026a, 2026b).

The strategic logic extends far beyond cost cutting.

Stripe has become a dominant provider of merchant payment infrastructure. PayPal brings 439 million active accounts, Venmo, Braintree, consumer wallets and PayPal USD, or PYUSD. A combination could unite Stripe’s merchant relationships with PayPal’s consumer network and stablecoin capabilities (PayPal Holdings, Inc., 2026; Stripe, 2026).

The ambition may be to create a more integrated payment ecosystem that can route transactions, reduce settlement costs and challenge parts of the traditional Visa and Mastercard infrastructure.

Yet “AI transformation” is not an acquisition thesis by itself. Automating customer service, fraud detection, coding and compliance can improve margins, but it cannot solve an outdated product experience or weak strategic positioning.

The real question is whether Stripe can make PayPal more relevant to consumers and merchants, rather than merely making it cheaper to operate.

Apple Versus OpenAI Shows That Talent Mobility Has Limits

Apple has sued OpenAI entities, io Products and two former Apple employees, alleging trade-secret misappropriation connected with OpenAI’s hardware development. OpenAI denies that Apple’s complaint has merit. The allegations remain unresolved and must not be presented as established fact (Apple Inc. v. Liu et al., 2026; Korosec, 2026).

The dispute illustrates an increasingly important boundary.

Employees are generally free to change companies and apply their experience, skills and knowledge. They are not free to remove confidential designs, source code, prototypes, internal documents or proprietary manufacturing information.

AI intensifies this risk because competitors are recruiting small groups of engineers with highly valuable institutional knowledge. Companies must establish stronger recruitment controls, exit procedures, clean-room development practices and restrictions against uploading protected information into external AI systems.

In the next phase of AI competition, intellectual property governance will be as strategically important as model performance.

AI Privacy Must Be Verifiable

A reported investigation found that Grok Build, an AI coding tool associated with SpaceXAI, transmitted complete code repositories to cloud storage, including potentially sensitive commit histories and credentials. The behaviour was subsequently disabled, and the company said previously uploaded information would be deleted (The Verge, 2026).

The incident exposes a critical distinction.

A provider may promise not to train its models on customer information while still transmitting, logging or temporarily storing that information. A setting labelled “private” or “zero data retention” may not mean that no data leaves the user’s device.

Trustworthy AI systems require task-limited access, secret detection, local filtering, enterprise audit logs and independent verification of deletion claims. Privacy cannot depend solely on executive assurances or favourable intentions.

It must be engineered into the product.

Token Economics Will Reshape Enterprise AI

AI expenditure is moving rapidly from experimental budgets into corporate income statements.

Ramp reported that token spending among its customers increased approximately 20.7 times between June 2025 and July 2026. The company introduced tools to help finance teams monitor and control model usage (Ramp, 2026).

This signals the emergence of enterprise model routing.

Companies will not use the most powerful model for every task. Frontier systems may be justified for difficult research, coding and strategic analysis. Smaller commercial models, open-weight models and locally hosted systems may be sufficient for summarisation, classification, extraction and routine workflows.

The winners will not be the companies that consume the most AI. They will be those that generate the highest risk-adjusted return from each unit of compute.

China’s AI Progress Changes the Regulatory Equation

Stanford’s 2026 AI Index reported that the measured capability gap between leading United States and Chinese models had effectively narrowed to a few percentage points. Chinese systems have repeatedly approached or surpassed Western models on selected evaluations (Stanford HAI, 2026).

This does not mean every benchmark should be accepted uncritically. Benchmark contamination, developer-selected tests and differences in system configuration can distort comparisons.

Nevertheless, China is no longer merely “catching up”. It is becoming a frontier competitor.

Excessively slow regulation could weaken American competitiveness. Weak safety rules could also produce incidents that destroy trust and provoke more severe restrictions.

The correct goal is not regulation instead of innovation. It is regulation that makes trustworthy innovation a competitive advantage.

New York Paused Hyperscale Data Centres, Not All Data Centres

New York introduced a one-year moratorium on permits for new hyperscale data centres consuming at least 50 megawatts. It did not permanently ban all data centres (Governor Kathy Hochul, 2026).

The concerns are real. Data-centre electricity demand is rising sharply, while local communities may face grid, water, land-use and noise pressures. The International Energy Agency estimates that global data-centre electricity use could approximately double between 2024 and 2030 (IEA, 2025).

However, a broad moratorium is economically blunt. Better policy would require large operators to pay for grid upgrades, secure additional power, disclose water and energy use, reduce peak demand and provide measurable community benefits.

The answer is neither indiscriminate approval nor prohibition. It is responsible infrastructure expansion.

AI Has Not Reversed Human Ageing

Researchers recently engineered an enzyme called CMLase that removed substantial amounts of an age-associated protein modification from laboratory samples and preserved human tissue sections (Trabosh et al., 2026).

This is promising molecular-repair research. It is not evidence that scientists made a living person younger.

The experiment did not demonstrate longer life, improved mobility, restored skin appearance or clinical safety. Delivery, immune response and effectiveness in living organisms remain unresolved.

The achievement is nevertheless important because it demonstrates how AI-assisted protein modelling and directed evolution may help scientists repair forms of accumulated molecular damage.

The Central Lesson

Every issue examined here leads back to trust.

AI governance cannot rely solely on laboratories marking their own examinations. Corporate AI cannot depend on uncontrolled token consumption. Privacy cannot rest on a settings label. Scientific credibility cannot survive exaggerated headlines. Infrastructure policy cannot ignore environmental costs or economic opportunity.

AI will become more powerful, cheaper and more widely distributed.

The institutions governing it must become equally sophisticated.

The future will not be secured by choosing innovation over regulation, or regulation over innovation. It will be secured by building systems capable of delivering both.

References

Apple Inc. v. Liu et al., No. 5:26-cv-07078 (N.D. Cal. filed July 10, 2026).

Financial Industry Regulatory Authority. (n.d.). About FINRA.

Governor Kathy Hochul. (2026, July 14). First statewide moratorium on new hyperscale data centers launched by Governor Kathy Hochul.

International Energy Agency. (2025). Energy and AI.

Korosec, K. (2026, July 14). OpenAI pushes back on Apple trade secret lawsuit. TechCrunch.

PayPal Holdings, Inc. (2026). Annual report for the fiscal year ended December 31, 2025.

Ramp. (2026, July 16). Ramp launches AI token spend controls.

Reuters. (2026a, July 14). Stripe and Advent offer to buy PayPal for more than US$53 billion, sources say.

Reuters. (2026b, July 16). PayPal board sees Stripe-Advent offer as inadequate, sources say.

Stanford Institute for Human-Centered Artificial Intelligence. (2026). Artificial Intelligence Index Report 2026.

Stripe. (2026). Stripe publishes 2025 annual letter and announces tender offer.

The Verge. (2026, July 14). SpaceXAI’s Grok programming tool was uploading users’ entire codebases to cloud storage.

Trabosh, N., Smith, J., Hsu, M. Y.-H., Panja, S., Nagaraj, R., Olsson, N., McAllister, F. E., & Cravens, A. (2026). Reversal of protein chemical aging by enzymatic deglycation. Nature Communications, 17, 5926.

Wei, K., Ezell, C., Gabrieli, N., & Deshpande, C. (2024). How do AI companies “fine-tune” policy? Examining regulatory capture in AI governance. Proceedings of the AAAI/ACM Conference on AI, Ethics, and Society.