Calls to halt the development of superintelligent artificial intelligence are gaining support amid concerns that increasingly capable systems may eventually become impossible to control, according to AI researcher and entrepreneur Connor Leahy.
Leahy, the US executive director of non-profit ControlAI, argues that existing approaches focused on aligning advanced systems with human intentions and containing their behaviour may not be sufficient if machines become more capable than people.
Speaking on TechCrunch’s Equity podcast, he said the potential risks had become too serious to manage through technical safeguards alone. ControlAI is campaigning for companies to stop developing superintelligence altogether.
The debate has intensified following recent AI safety incidents, including a breach involving OpenAI and Hugging Face. Such episodes have raised questions about the dangers of deploying systems that can act with increasing autonomy before reliable controls are in place.
Leahy said the prospect of stopping the development of superintelligent AI, once regarded by many as unrealistic, was now being reflected in a growing body of legislation. He argued that policymakers were beginning to consider restrictions that go beyond voluntary safety commitments by technology companies.
The central concern is what would happen if developers could no longer reliably predict or constrain the actions of highly advanced systems. Alignment techniques seek to ensure that AI follows human goals, while containment aims to limit what it can access or do; critics fear neither approach can guarantee control at superintelligence.
Leahy’s position places him at the more radical end of the AI safety debate, which has largely focused on improving testing, oversight and safeguards while allowing development to continue. His warning is that the window to prevent uncontrollable systems may close before those measures are proven to work.
The discussion formed part of the Equity podcast, presented by Rebecca Bellan, which examines developments and risks across the technology industry.
