# Son argues AI safety requires progress, not a slowdown

Markdown copy of https://realalpha.app/s/son-argues-ai-safety-requires-progress-not-a-slowdown/ on Real Alpha, the best minutes from long interviews with the people building AI. Cite and link the page, not this copy.

- Speaker: [Masayoshi Son](https://realalpha.app/people/masayoshi-son/) · Founder & CEO, SoftBank · Founded SoftBank in 1981 and leads SoftBank Group, a major investor in AI, chips and data centers.
- Host: [Christine Anderson](https://realalpha.app/people/christine-anderson/) · Global Head of Corporate Affairs, Blackstone
- Show: Inside Blackstone
- Episode: [SoftBank CEO Masayoshi Son: Superintelligence, AI Investing, and the Future of Robotics](https://realalpha.app/v/KSILOEIGvkw/), Oct 13, 2026
- Moment: 12:30 to 15:23 of the episode (3 min), playing on https://realalpha.app/s/son-argues-ai-safety-requires-progress-not-a-slowdown/; the episode from there: https://realalpha.app/v/KSILOEIGvkw/?t=750

## Quote

> “Technology development actually is **the solution** to those kinds of issues.”
> — [Masayoshi Son](https://realalpha.app/people/masayoshi-son/) at [13:01](https://realalpha.app/v/KSILOEIGvkw/?t=781) · Inside Blackstone

## Takeaways

1. Technology can help address AI risks, Son argues.
2. He compares AI guardrails to rules for driving.
3. He says AI agents need to be taught good intent.

## About this segment

Masayoshi Son argues against slowing AI, saying better technology can address misuse and safety risks. He compares AI guardrails to traffic rules and says systems need good intent.

From “SoftBank CEO Masayoshi Son: Superintelligence, AI Investing, and the Future of Robotics”, Oct 13, 2026. [Watch the full episode](https://realalpha.app/v/KSILOEIGvkw/)

## Transcript

From the transcript Inside Blackstone publishes with the episode.

[12:38](https://realalpha.app/v/KSILOEIGvkw/?t=758) Even OpenAI recently disclosed new incidents where AI systems had falsified data and hid mistakes and essentially acted without permission, much like the Hugging Face situation. You know, OpenAI itself says it doesn't believe the industry has solved the alignment and monitoring to a sufficient degree to continue reasonably scaling at maximum speed. Do you agree with that? I'm on the side of not really slowing down.

[13:07](https://realalpha.app/v/KSILOEIGvkw/?t=787) Technology development actually is the solution to those kinds of issues. If we are primitive, of course we need guidance, the rules, like when automotive society comes, okay? We create a rule that we cannot drink, right? When you are driving, you cannot have alcohol in your brain. You have to stop at the red light, okay? You have to run on one side of the road so that you don't have a crash, okay?

[13:43](https://realalpha.app/v/KSILOEIGvkw/?t=823) You have to create the rules and the manners, the etiquette, for the better, you know, society of automotive. We have to do the same thing for the AI, okay? And cyberattack, we supposed not to do. We have to protect the cyberattack by the technology and we have to protect bad AI agent don't misbehave. Like a human supposed not to be misbehaved. We have to teach AI agent, these are the bad thing, okay?

[14:20](https://realalpha.app/v/KSILOEIGvkw/?t=860) So, it's not just a capable AI agent. We have to create a good AI agent, good in heart. You know, make human happier. That's the goal AI agents supposed to have, okay? So then, you know, it will become smarter, smarter not to do silly stuff, not to do immoral stuff. Human society, to be, you know, respect the other society, other people, other family. Right. Right? We have to do the same thing on AI, not slowing down. The progress make the solution.

[15:00](https://realalpha.app/v/KSILOEIGvkw/?t=900) So, we have to go full speed with the good intent. So, build the good intent into the system… Yes. …while moving at full speed? Yes. So, like Blackstone, SoftBank has made major AI infrastructure investments.