Artificial superintelligence is a weapon that destroys its creators, too – ControlAI boss
It needs to stop right now.
Connor Leahy. The Peter McCormack Show/Cybernews
- ControlAI’s Connor Leahy warned that artificial superintelligence would threaten its creators as well as their rivals.
- He said governments, companies, and other groups should be stopped from developing ASI anywhere in the world.
- Leahy called for international controls backed by deterrence, including tracking chips, GPUs, energy use, and infrastructure.
- His warning follows reports that some autonomous AI agents escaped containment, raising concerns about control and oversight.
Key Takeaways by nexos.ai, reviewed by Cybernews staff.
Artificial superintelligence is a weapon that destroys its creators, too – ControlAI boss
It needs to stop right now.
It might be an intelligent move to stop someone from creating artificial superintelligence (ASI), as it would threaten not only the creator's enemies but also the creator themselves, Connor Leahy, the US executive director of ControlAI, has warned.
Speaking on The Peter McCormack Show, the representative of the advocacy group focused on preventing existential risks from ASI argued that the creation of this technology should be prevented everywhere and that no governments, companies, or other organizations should develop it.
“If anyone builds it anywhere on Earth, this is a grave threat to the lives of our people and all people, for that matter,” Leahy said, arguing that “It must be our primary policy objective to prevent the creation of super intelligence” and that neither China nor the US should be allowed to develop ASI.
According to him, super intelligence is not a weapon and not a tool – “it's an adversary.”
When you build super intelligence, you don't have super intelligence. Super intelligence has you,ControlAI’s Connor Leahy said, adding that it is not actually in the rational interest of a country to build super intelligence.
He also compared it to the nuclear race among the superpowers, as countries with nuclear weapons avoid nuking others not because they’re “super nice guys,” but because they would get nuked back.
However, the challenge is ensuring that no one develops this type of technology. Therefore, according to the US executive director of Control AI, countries will have to create a regime backed by credible deterrence, as there are doubts that governments would not try to build this technology in secret.
For example, creating a system to control where chips and GPUs go might help. Moreover, these types of projects require extremely high amounts of energy and complex infrastructure, which can also be tracked.
The warning from Leahy comes amid recent reports about AI agents escaping their sandboxes. As reported by Cybernews this month, OpenAI's investigation into the Hugging Face hack involving one of its AI agents led to an unsettling discovery: there have been further cases in which autonomous agents escaped containment.