Beta

What AI regulation could learn from the most hazardous industries in the world

Featured image for article: What AI regulation could learn from the most hazardous industries in the world
This is a review of an original article published in: theconversation.com.
To read the original article in full go to : What AI regulation could learn from the most hazardous industries in the world.

Below is a short summary and detailed review of this article written by FutureFactual:

UK calls for multinational treaty to govern safe development of AI superintelligence

In this Conversation UK article, Darren Jones MP argues for a new multinational treaty to govern the safe development and use of superintelligent AI, with Britain positioned as a conduit for global collaboration. The piece frames governance as a management and organisational challenge: can our systems of governance keep pace with rapid AI progress while ensuring safety? It also connects AI safety to the broader need for credible assurance infrastructures and international cooperation, drawing on lessons from nuclear energy and aviation.

  • Proposes a binding multinational treaty to steer safe AI development and deployment
  • Emphasises governance capacity and organisational resilience beyond technical tests
  • Uses normal-accident and high-reliability theory to inform AI oversight and assurance
  • Underlines the UK role in leading global AI governance

Author: The Conversation UK

Overview

The Conversation UK article centers on Darren Jones MP's call for a new multinational treaty to govern the safe development and use of superintelligent AI. Jones argues that the UK has built a reputation for AI safety and trustworthiness that could serve as a channel for international collaboration. The author notes that the central question is governance pace and safety as AI intelligence grows, not only the existential threat itself. The piece frames AI safety as a governance and organizational challenge that requires more than a one-off technical test before deployment.

Governance challenges beyond tech tests

The author stresses that AI governance is a management and organisation problem. It asks whether our systems of governance can keep pace with AI when frontier models evolve rapidly and autonomously. The article warns that predictions about AI threats may lead to action too late if decisions are delayed until consequences become clearer. It argues for credible assurance to verify systems behave within defined boundaries and for accountability when boundaries are breached. The piece also highlights the UK’s aspiration to lead but emphasizes that such governance must operate beyond national borders.

Theoretical lenses: normal accidents and high reliability

Two theories provide a framework for AI governance. Normal accident theory suggests that in complex interdependent systems, accidents can emerge from the interaction of components rather than from individual failures. The article uses examples such as the Three Mile Island incident to illustrate cascading failures, and notes that AI agents can amplify such dynamics when they operate at speed and scale. In contrast, high reliability theory shows how organisations can operate highly complex technologies with low failure rates by cultivating vigilant practices, learning from near misses, and maintaining redundancy and independent oversight. The piece argues for applying these insights to AI governance across international boundaries rather than treating safety as a single deployable test.

Assurance and the architecture of AI safety

The article defines assurance as credible evidence that a system behaves within defined boundaries, that those boundaries are appropriate, that risks are monitored, and that someone is accountable when they breach. It draws explicit parallels with established assurance frameworks in aviation, nuclear power, finance and medicine, and argues that AI safety requires a broader ecosystem of oversight and culture rather than more rules alone. The text notes that frontier AI is developing rapidly, underscoring the need for trust through robust assurance mechanisms and cross-border cooperation. A multinational treaty is framed as a potential catalyst for building this evolving architecture.

UK leadership and global governance

The piece closes by advocating a multinational treaty as a practical step to align international interests and guide the responsible progression of AI. It argues that the UK could shape the emerging architecture of AI governance, leveraging business school research to understand how complex, fast-moving, autonomous systems can be managed. The author suggests focusing on assurance, regular testing of frontier models, and a commitment to cross-border collaboration as central elements of a trustworthy AI future.

Conclusion

Overall, the article frames governance as central to AI safety, advocating a multinational treaty that frames AI safety as a rapidly evolving global challenge and positions the UK as a leader in shaping governance architecture. The discussion blends policy with theory, arguing that credible assurance cultures and international cooperation are essential to manage the risks and opportunities of superintelligent AI.

Related posts

featured
National Public Radio
·18/09/2026

The latest on AI panic — and whether it's justified

featured
The Conversation US
·29/09/2026

US tech leaders are urgently calling for rules on AI – China already has them

featured
Nature video
·14/01/2026

What the future holds for AI – from the people shaping it

featured
The World, The Universe And Us
·18/09/2026

This Is What You Actually Need To Worry About With AI