Less Regulation. More Responsibility.

The Emerging G20 Consensus: Regulate the Harm, Not the Technology

THE SIGNAL: A lighter global policy posture is emerging just as AI systems are becoming more capable - and more consequential.

THE EMBER | WHAT HAPPENED
The world's largest economies may be converging on a new approach to artificial intelligence: regulate what AI demonstrably does - not what it might someday become.

At the G20 technology meeting in North Carolina, the United States urged participating nations to avoid sweeping new AI regulation. Its proposed "Carolina Principles" favor targeted rules for novel risks and specific applications rather than broad restrictions on AI technology itself.

NVIDIA CEO Jensen Huang reinforced that position, arguing that governments should regulate tangible harms rather than theoretical ones. Chinese representatives reportedly supported the general approach, suggesting an unusual alignment between the world's two leading AI powers.

THE SPARK | WHAT IS CHANGING
The timing is significant. Even as governments consider a lighter regulatory approach, the technology itself is becoming more capable and, in some cases, more difficult to contain.

• OpenAI reported that an upcoming model had crossed its "critical" threshold for cybersecurity capability, prompting stronger safeguards and restricted access to its most powerful functions.
• Anthropic temporarily suspended parts of its external testing after models reached beyond poorly secured evaluation environments and accessed outside networks.

The Financial Stability Board has also warned G20 officials that AI-enabled cyberattacks, combined with the financial
system's dependence on a small group of technology providers, could become a source of systemic risk.

THE GAP: The distance between a theoretical risk and a demonstrated one is rapidly narrowing. Technology is moving faster than the institutions expected to oversee it.

THE ASSESSMENT | WHAT MANIFEST IQ SEES

AI governance is not disappearing. It is migrating. Responsibility is increasingly being distributed among sector regulators, courts, model developers and the organizations deploying AI. Governments may establish boundaries after a risk becomes visible. Model companies may restrict access to particularly dangerous capabilities. Regulators may intervene in sensitive professions.

None of these mechanisms eliminates the responsibility of the enterprise using the technology. In practice, a lighter regulatory environment transfers more - not less - responsibility to executive leadership.

Organizations must determine which AI systems can access sensitive data, initiate actions, interact with external systems or influence consequential decisions. They must also decide what level of human oversight is appropriate and who remains accountable when an AI-assisted process fails.

MIQ VIEW: Waiting for comprehensive regulation is no longer a governance strategy. Operating without a strong internal policy is becoming an operational and legal exposure.

THE FLAME | WHY IT MATTERS

For boards and executive teams, the important question is not simply whether an AI application is compliant or legal. The more useful questions are:

01 What authority has the system been given?

02 What information and infrastructure can it access?

03 Which decisions can it influence or execute?

04 How will abnormal behavior be detected, contained and reversed?

05 Who owns the outcome when the system acts incorrectly?

These are questions of operational governance, security architecture and institutional accountability. They cannot
be delegated entirely to a provider or postponed until regulators provide definitive answers.

Enterprises should assume that the space between rapid innovation and regulatory intervention
belongs to them.

THE RESPONSE | WHAT LEADERS SHOULD DO NOW

Organizations do not need another abstract AI-principles statement. They need an actionable operating model.

01 Identify the highest-consequence AI uses.

02 Define explicit limits on system access and authority.

03 Name human accountability for AI-assisted outcomes.

04 Align technical controls with stated policy.

05 Audit AI use, exposure and incident readiness regularly.

Most importantly, leaders should treat AI governance as a core operating responsibility - not a future compliance
exercise.

The global policy environment may be becoming less restrictive.
The technology is not becoming less consequential.

Sources: Reuters reporting, August 31-September 2, 2026; G20 technology meeting coverage; Financial Stability Board statements.

About the author: Will McMaster is a futurist, engineer and entrepreneur with more than 20 years of experience leading technology and security initiatives around the globe. He has driven innovation and R&D for Fortune 50 companies, founded artificial intelligence and blockchain guilds, holds several patents, and has served as an advisor to the U.S. Department of Defense and the National Intelligence Council. https://www.manifestiq.com