Public transcript

AI Industry Has Done 'Terrible Job' of Communicating Benefits, Says Sam Altman

By: Bloomberg Podcasts
Length: 13 min
Language: English
Text from: the video's captions
Transcribed: 13 October 2026

Read the full transcript Open on YouTube

Notes in English.

Summary

Sam Altman, CEO of OpenAI, speaks with Bloomberg about the release of Astra, which he frames as an early step toward AGI and the first model everyday users can reliably hand complex tasks like building software. He explains the cyber critical safeguards required before release, the tiered cyber access rollout, and the importance of chain of thought monitoring. He also agrees the AI industry has done a terrible job explaining AI's benefits, discusses aggressive price cuts, and says OpenAI would remain mission first if it ever goes public.

Key points

Sam Altman says Astra feels like a new step toward models that help people create value, do work, and discover new science.
Astra hit a 'cyber critical' threshold, requiring new safeguards under OpenAI's preparedness framework before release.
OpenAI is rolling out Astra with tiered cyber access, starting with Trusted Access Partners, because models like Astra are needed to defend against a coming wave of AI-enabled cyber attacks.
Altman says the model being launched is GPT six, and the previously paused model was a future model.
Altman acknowledges the industry, including OpenAI, has done a terrible job explaining AI's benefits to ordinary people.
He argues AI should give people more power and autonomy, and that a cure for cancer alone is not a high enough ambition.
OpenAI recently cut the price of Luna, its small model, by 80% and plans to keep dropping prices dramatically.
If OpenAI goes public someday, Altman plans to emphasize that the company is mission first and extremely long term oriented.

Questions it answers

00:00Astra's new capabilities

What makes Astra different from earlier models for everyday users?

Altman says Astra feels like a new step toward models that create value, and it is the first model he would tell everyday users to just try, with use cases like games, DIY electrical projects, simulations, and financial models.

First model where everyday users can reliably build complex software
Use cases span games, DIY engineering, science simulations, and financial modeling

02:11Guardrails and cyber access

Why was Astra released with specific guardrails?

Astra hit cyber critical under the preparedness framework, requiring new safeguards, and is being released with tiered cyber access starting with Trusted Access Partners, because Altman believes models like Astra are needed to defend against a coming wave of AI-enabled cyber attacks.

Astra hit cyber critical and needed new safeguards
Tiered cyber access starts with Trusted Access Partners
The launching model is GPT six; the paused one was a future model

04:23Chain of thought monitoring

Can OpenAI see what Astra is planning before it acts?

OpenAI emphasizes monitorability including chain of thought monitoring, even at the cost of some capabilities, alongside sandboxing and alignment as part of defense in depth.

Chain of thought monitoring is preserved even if it limits capabilities
Monitorability is one part of a broader defense in depth

06:29Red lines and autonomy

Would OpenAI detect and shut down a dangerous model?

Altman says there are clear red lines the company can detect, but society will have to negotiate acceptable use, comparing AI to electricity as a platform everyone has a right to use.

Some clear red lines exist and are detectable
Acceptable use will require ongoing societal negotiation

07:53Industry messaging failure

Has the AI industry communicated its benefits well?

Agreeing with Treasury Secretary Scott Besson, Altman says the industry, including OpenAI, has done a terrible job, and wants AI to give people more power and autonomy rather than just promising a cure for cancer.

Altman admits the industry and OpenAI have done a bad job
He wants a boom in creativity and entrepreneurship, not less user power

09:58Pricing and access

How is OpenAI balancing pricing and access?

OpenAI wants incredibly capable, low-cost, abundant intelligence, recently cut Luna's price by 80%, and plans to keep dropping prices dramatically while still making money.

Luna's price was cut by 80%
Goal is best price performance at every level of intelligence

11:24Public vs. private

Would OpenAI stay private or go public?

Altman assumes OpenAI will go public someday and plans to remind investors the company is mission first, extremely long term oriented, and will make decisions on a multi-decade timeframe for society as a whole.

Altman assumes OpenAI will go public someday
Mission and multi-decade thinking would come first

Notes

Astra's capabilities

Sam Altman describes Astra as an early step toward AGI and a new step toward models that help create value, do work, and discover new science.
It is the first model where he would tell an everyday person to "just give it a try" with complex software tasks.
Example uses he has watched: computer games, home DIY electrical engineering projects, complex scientific simulations, financial models with interactive code.

Safety and guardrails

Astra hit "cyber critical" under OpenAI's preparedness framework, requiring a new set of safeguards before release.
The model launching is GPT six; the recently paused model was a future model.
Cyber access is tiered: rolled out first to Trusted Access Partners, then more broadly in coming days.
Altman argues the world is close to a complete change in the cyber attack landscape, and tools like Astra are needed to collectively defend against new threats.

Monitoring and control

OpenAI emphasizes monitorability, including chain of thought monitoring, even at the cost of not maximizing capabilities.
Defense in depth also includes sandboxing and alignment.
Altman says there are clear red lines for what models should not do, but society will have to negotiate acceptable use, comparing AI to electricity as a platform that belongs to everyone.

Industry messaging

Responding to Treasury Secretary Scott Besson's G20 remark, Altman agrees the industry has done a terrible job explaining AI's benefits, and says OpenAI has done a bad job too.
He wants AI to enable a boom in creativity and entrepreneurship, giving people more power and autonomy, not less; "I don't think a cure for cancer is enough."

Pricing and company structure

OpenAI cut the price of Luna, its small model, by 80%, and aims for the best price performance at every level of intelligence.
If OpenAI goes public someday, Altman plans to remind investors the company is mission first and extremely long term oriented, deciding on a multi-decade timeframe.

Text from the video's own captions. Summary and notes written by AI from the transcript, so check anything important against the recording.