Public transcript

Altman on OpenAI's IPO Plans, AI Agent Dots, AI Security

By: Bloomberg Television
Length: 10 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 speaks with Bloomberg's Ed Ludlow at OpenAI's Dev Day about the launch of the always-on agent Dots, which he describes as a new category of product aimed largely at enterprises. He explains why OpenAI held back the release of Astra 6.1 and paused training on another model, saying safety, alignment and security work justified the delay. He also discusses IPO timing, saying OpenAI wants to be public one day but not while adjusting to new capability levels, and calls for shared industry safety standards.

Key points

Altman says Dots is largely an enterprise product for getting work done, with a mass-market consumer version to come later.
He describes Dots as a completely new category of agent in terms of capability and compute, priced at the top end of the market.
OpenAI held back the release of Astra 6.1 and paused training on a separate model to put more attention into safety, alignment, monitoring and security.
Altman declined to comment on the reported $70 billion ARR figure but said the company is in a very steep growth period.
He says OpenAI wants to be public one day, but being a newly public company while adjusting to new capability levels would add incompatible pressure.
Altman calls for shared standards for safety cases and shared methods, such as independent evaluators or government review.
He says disclosing existential risk has nothing to do with going public and is simply OpenAI's responsibility to its mission.

Questions it answers

00:00Introducing Dots

Who is OpenAI's new agent Dots for?

Altman says Dots is largely an enterprise product for getting work done, built on the best model, Astra, with a mass-market consumer version to come over time.

Dots is aimed at the top end of the market, mainly enterprise.
A mass-market consumer version will come later.

01:35Pricing and positioning

How does Dots differ from other agents on the market?

He says it is a completely new category in capability and compute, and while priced higher, OpenAI's strategy is to lead the Pareto frontier by driving quality up and price down.

Altman has tried competitor agents but doesn't use them regularly.
OpenAI aims to lead the Pareto frontier of quality versus price.

03:23Business momentum

How is OpenAI's business performing?

He declined to comment on the reported $70 billion ARR figure but said the company is in a very steep growth period across consumers, developers and enterprises.

Altman won't confirm leaked numbers.
ChatGPT is growing phenomenally well.

04:32Holding back Astra 6.1

Why did OpenAI hold back Astra 6.1 and pause training on another model?

He says OpenAI is entering a steep part of the capabilities curve and wants to put more attention into safety, alignment, monitoring and security, while still releasing faster and cheaper versions of existing models.

OpenAI will slow training or releases for safety.
Faster and cheaper versions of existing models can still ship.

05:54Speed, price and capability

Why release speed and price improvements instead of just more capable models?

He says people want all three, and ultrafast in particular is special and hard to give up, though it is more expensive to run and priced accordingly.

Ultrafast is more expensive to run but OpenAI charges more.
Altman finds ultrafast hard to go back from.

07:11IPO timing

What did Altman say about an OpenAI IPO?

He wants to be public one day, but being newly public while adjusting to new capability levels and safety requirements would add incompatible pressure; investors are described as patient.

OpenAI will go public at some point, just not now.
Investors are happy and patient.

07:59Shared safety standards

What collective AI safety effort does Altman want to see?

He proposes shared standards for safety cases and shared verification methods such as independent evaluators or government review, and says disclosing existential risk is a mission responsibility unrelated to going public.

Shared safety standards and verification methods are needed.
Risk disclosure is framed as OpenAI's responsibility to its mission.

Notes

The Dots agent

Altman describes Dots as a very capable, always-on agent built on OpenAI's best model, Astra.
It is largely an enterprise product for getting work done, though many prosumers building startups will use it; a mass-market consumer version will come over time.
Staff interact with Dots in a conversational style that differs from traditional ChatGPT, without quite anthropomorphizing it.
Altman has tried competitors' agents but has not found a regular place for them in his life; he calls Dots a completely new category.

Pricing and strategy

Dots is priced higher than other products because of its capability and the compute behind it.
OpenAI's goal is to drive quality up and price down, leading the Pareto frontier at every point, because better AI drives more usage.
Ultrafast is more expensive to run but OpenAI charges more for it; Altman says the market will decide.

Business momentum

Altman declined to comment on the reported $70 billion ARR third-quarter number but said growth is very steep across consumers, developers and enterprises.

Safety and model releases

OpenAI held back Astra 6.1 and paused training on a separate model to focus more on safety, alignment, monitoring and security.
Altman says the world should have confidence in OpenAI's safety cases and that it will act responsibly without waiting for regulation or competitors.
Faster and cheaper versions of existing models can still be released reasonably.

IPO plans

Altman wants OpenAI to be public one day, but being newly public while adjusting to new capability levels would add incompatible pressure; investors are described as happy and patient.

Collective safety effort

He proposes shared standards for safety cases and shared methods of verification, via independent evaluators, government review, or companies checking each other's work.
Disclosing existential risk is framed as a responsibility of OpenAI's mission, unrelated to going public.

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