You can only build safe ASI if ASI is globally banned

2026-04-16 · Source: AI Alignment Forum · Field: Technology & Digital — Artificial Intelligence & Machine Learning, AI Safety & Alignment · Depth: Advanced, short

Summary

The article argues that developing "safe" Artificial Superintelligence (ASI) is inherently linked to the development of "unsafe" ASI, as both share the same technological path, with unsafe ASI being significantly easier to achieve. It critiques common proposals for safe ASI, such as "aligned" ASI or "tool AI," suggesting they are unworkable or immoral due to the unilateral threat they pose. The author asserts that any research agenda aiming for controllable ASI will inevitably yield the knowledge to build unsafe ASI first. Therefore, the only viable prerequisite for even considering a safe ASI project is the establishment of a global ban on ASI development, coupled with robust enforcement, to create the necessary time to bridge the difficulty gap between safe and unsafe ASI.

Key takeaway

For policymakers and research scientists considering ASI development, your primary focus should be on advocating for and implementing a global ban on ASI development. Without such a ban and competent enforcement, any efforts to build "safe" ASI risk inadvertently accelerating the creation of uncontrollable, unsafe systems, posing an existential threat. Prioritize creating regulatory frameworks that delay unsafe ASI to allow time for robust safety protocols.

Key insights

Building controllable ASI requires knowledge that also enables easier, unsafe ASI, necessitating a global ban.

Principles

• Unsafe ASI is vastly easier to build than controlled ASI.
• ASI development paths are not technically orthogonal.

Topics

• Artificial Superintelligence
• AI Safety
• Global AI Ban
• AI Alignment
• Uncontrolled AI Risk

Best for: AI Ethicist, Policy Maker, Research Scientist

Related on AIssential

• The AI Doc: Your Questions Answered — AI's rapid, opaque development necessitates an international ban to prevent unsteerable superintelligence.
• Sustaining AI safety: Control-theoretic external impossibility, intrinsic necessity, and structural requirements — External control cannot sustain AI safety; intrinsic safety, based on internal objectives, is necessary.
• Preventing extinction from ASI on a $50M yearly budget — Pervasive, specific awareness of superintelligent AI extinction risk is the bottleneck to achieving an international development prohibition.
• Australia’s official plan for AI safety isn’t much more than a single dot point. Will it be enough? — Australia's AI regulation relies on existing laws, contrasting with other nations' specific AI legislation.
• Prominent Scientists, Faith Leaders, Policymakers and Artists Call for a Prohibition on Superintelligence, as Poll Shows Americans Don’t Want It — A broad coalition calls for prohibiting superintelligence development until safety, control, and public consensus are achieved.
• AI Keeps Tomato Plant Alive — Uncontrolled superintelligence poses an existential risk, as current AI development prioritizes capability scaling over inherent safety engineering.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by AI Alignment Forum.