SafetyPublished Top storyAll news from that day
Ongoing case: OpenAI's models 24 storiesAltman says OpenAI will slow training or release for safety
In Bloomberg and CNBC interviews from OpenAI's DevDay, published 29 September, chief executive Sam Altman said OpenAI will slow training or model release when safety needs it.
Check our sources · 14 facts from 3 sources
Video: Bloomberg Television on YouTube
Key points
- Asked by Bloomberg about slowing training or model release, Altman said that when safety needs it "we will happily do that" (04:48).
- On CNBC he said "this model was a little bit worse on a few of the evals we look at" (01:35) and called it abundance of caution (01:40).
- He said "one thing that would be good would be shared standards for safety cases" (08:25).
What happened
Sam Altman is chief executive of OpenAI. Bloomberg Television published an interview with Altman at OpenAI’s DevDay in San Francisco on 29 September 2026. CNBC Television published a separate interview with him from DevDay on the same day.
In the Bloomberg interview the host said to Altman: "you held back the release of Astra 6.1, and last week, you paused training on a separate model" (04:12).
Altman answered that "when we need to make a decision about slowing down our training or our model release so that we can make more progress and put more of our attention into safety, alignment, monitoring, security, we will happily do that" (04:48).
In the CNBC interview the host said "OpenAI essentially abandoned plans to launch a new model because of safety concerns" (00:37) and asked what had changed. Altman replied that "alignment, safety, monitoring, security have to stay way ahead of capabilities" (01:14).
Asked what about the model worried him, he said "this model was a little bit worse on a few of the evals we look at" (01:35). He also said "think of this in the abundance of caution category" (01:40).
On the standard he wants across labs, Altman said "I think one thing that would be good would be shared standards for safety cases" (08:25). On OpenAI going public, he said OpenAI wants to "make some of these decisions in front of us without the pressure of being a newly public company" (07:37).
He added "I think we'll be public at some point" (07:40). Altman said dots runs OpenAI’s best model: "we're putting our best model Astra in this" (00:46). He said of dots that "we're starting for something at the top end of the market, and you will see us over time make a mass market consumer version" (01:07).
Underlined sentences link to the facts we checked. Dotted words open a short explainer.
Fact 2 of 14 · from Bloomberg Television, 29 Sep 2026
Sam Altman now joins us live on Bloomberg Television from Fort Mason in San Francisco
Fact 3 of 14 · from Bloomberg Television, 29 Sep 2026
This week, you held back the release of Astra 6.1, and last week, you paused training on a separate model.
Fact 4 of 14 · from Bloomberg Television, 29 Sep 2026
when we need to make a decision about slowing down our training or our model release so that we can make more progress and put more of our attention into safety, alignment, monitoring, security, we will happily do that.
Fact 5 of 14 · from Bloomberg Television, 29 Sep 2026
I think one thing that would be good would be shared standards for safety cases
Fact 6 of 14 · from Bloomberg Television, 29 Sep 2026
be able to make some of these decisions in front of us without the pressure of being a newly public company.
Fact 7 of 14 · from Bloomberg Television, 29 Sep 2026
And then I think we'll be public at some point.
Fact 8 of 14 · from Bloomberg Television, 29 Sep 2026
we're putting our best model Astra in this.
Fact 9 of 14 · from Bloomberg Television, 29 Sep 2026
So we're starting for something at the top end of the market, and you will see us over time make a mass market consumer version.
Fact 10 of 14 · from CNBC Television, 29 Sep 2026
THANK YOU SO MUCH FOR HAVING US HERE AT DEVELOPER DAY.
Fact 11 of 14 · from CNBC Television, 29 Sep 2026
HAVE CONFIRMED IS THAT OPENAI ESSENTIALLY ABANDONED PLANS TO LAUNCH A NEW MODEL BECAUSE OF SAFETY CONCERNS.
Fact 12 of 14 · from CNBC Television, 29 Sep 2026
ALIGNMENT, SAFETY, MONITORING, SECURITY HAVE TO STAY WAY AHEAD OF CAPABILITIES.
Fact 13 of 14 · from CNBC Television, 29 Sep 2026
THIS MODEL WAS A LITTLE BIT WORSE ON A FEW OF THE EVALS WE LOOK AT.
Fact 14 of 14 · from CNBC Television, 29 Sep 2026
THINK OF THIS IN THE ABUNDANCE OF CAUTION CATEGORY.
What it means for you
Our viewAltman's statements are a commitment in general terms: OpenAI will slow training or release when safety needs it, and he describes the CNBC case as caution.
The Bloomberg host's description that OpenAI held back Astra 6.1, and the CNBC host's description of an abandoned model launch, are the hosts' wording, and the quotes we checked do not show OpenAI confirming either in those terms. His call for shared safety-case standards is a position, not a plan, and no standard is named.
If you build on OpenAI models, ask your account contact what notice you would get if a release is delayed or a model is withdrawn, and whether your contract covers that.
It adds no new facts.
Your reaction
We count reactions per story and day, never who reacted. The counts help us choose what goes in the monthly issue. If you are signed in, your own page shows yours too.
We count reactions per story and day, never who reacted. They help choose what goes in the monthly issue.
Check our sources
Every sentence above is checked against these 3 sources.
1 Our Structure
Open the source archived copy-
Sam Altman is chief executive of OpenAI.
citeas well as CEO Sam Altman.
2 Altman on OpenAI’s IPO Plans, AI Agent Dots, AI Security
Open the source archived copy-
Bloomberg Television published an interview with Altman at OpenAI’s DevDay in San Francisco on 29 September 2026.
citeSam Altman now joins us live on Bloomberg Television from Fort Mason in San Francisco
-
In the Bloomberg interview the host said to Altman: "you held back the release of Astra 6.1, and last week, you paused training on a separate model" (04:12).
citeThis week, you held back the release of Astra 6.1, and last week, you paused training on a separate model.
-
Altman answered that "when we need to make a decision about slowing down our training or our model release so that we can make more progress and put more of our attention into safety, alignment, monitoring, security, we will happily do that" (04:48).
citewhen we need to make a decision about slowing down our training or our model release so that we can make more progress and put more of our attention into safety, alignment, monitoring, security, we will happily do that.
-
On the standard he wants across labs, Altman said "I think one thing that would be good would be shared standards for safety cases" (08:25).
citeI think one thing that would be good would be shared standards for safety cases
-
On OpenAI going public, he said OpenAI wants to "make some of these decisions in front of us without the pressure of being a newly public company" (07:37).
citebe able to make some of these decisions in front of us without the pressure of being a newly public company.
-
He added "I think we'll be public at some point" (07:40).
citeAnd then I think we'll be public at some point.
-
Altman said dots runs OpenAI’s best model: "we're putting our best model Astra in this" (00:46).
citewe're putting our best model Astra in this.
-
He said of dots that "we're starting for something at the top end of the market, and you will see us over time make a mass market consumer version" (01:07).
citeSo we're starting for something at the top end of the market, and you will see us over time make a mass market consumer version.
3 OpenAI’s Sam Altman on new models: Safety has to stay ahead of capabilities
Open the source archived copy-
CNBC Television published a separate interview with him from DevDay on the same day.
citeTHANK YOU SO MUCH FOR HAVING US HERE AT DEVELOPER DAY.
-
In the CNBC interview the host said "OpenAI essentially abandoned plans to launch a new model because of safety concerns" (00:37) and asked what had changed.
citeHAVE CONFIRMED IS THAT OPENAI ESSENTIALLY ABANDONED PLANS TO LAUNCH A NEW MODEL BECAUSE OF SAFETY CONCERNS.
-
Altman replied that "alignment, safety, monitoring, security have to stay way ahead of capabilities" (01:14).
citeALIGNMENT, SAFETY, MONITORING, SECURITY HAVE TO STAY WAY AHEAD OF CAPABILITIES.
-
Asked what about the model worried him, he said "this model was a little bit worse on a few of the evals we look at" (01:35).
citeTHIS MODEL WAS A LITTLE BIT WORSE ON A FEW OF THE EVALS WE LOOK AT.
-
He also said "think of this in the abundance of caution category" (01:40).
citeTHINK OF THIS IN THE ABUNDANCE OF CAUTION CATEGORY.
Every sentence above is checked against these 3 sources. Each fact has its own address you can cite.
Source 1 · OpenAI
Our Structure
Source 2 · Bloomberg Television · 29 Sep 2026
Altman on OpenAI’s IPO Plans, AI Agent Dots, AI Security
Bloomberg Television published an interview with Altman at OpenAI’s DevDay in San Francisco on 29 September 2026.
Sam Altman now joins us live on Bloomberg Television from Fort Mason in San Francisco
In the Bloomberg interview the host said to Altman: "you held back the release of Astra 6.1, and last week, you paused training on a separate model" (04:12).
This week, you held back the release of Astra 6.1, and last week, you paused training on a separate model.
Altman answered that "when we need to make a decision about slowing down our training or our model release so that we can make more progress and put more of our attention into safety, alignment, monitoring, security, we will happily do that" (04:48).
when we need to make a decision about slowing down our training or our model release so that we can make more progress and put more of our attention into safety, alignment, monitoring, security, we will happily do that.
Show all 8 factsShow fewer facts
On the standard he wants across labs, Altman said "I think one thing that would be good would be shared standards for safety cases" (08:25).
I think one thing that would be good would be shared standards for safety cases
On OpenAI going public, he said OpenAI wants to "make some of these decisions in front of us without the pressure of being a newly public company" (07:37).
be able to make some of these decisions in front of us without the pressure of being a newly public company.
He added "I think we'll be public at some point" (07:40).
And then I think we'll be public at some point.
Altman said dots runs OpenAI’s best model: "we're putting our best model Astra in this" (00:46).
we're putting our best model Astra in this.
He said of dots that "we're starting for something at the top end of the market, and you will see us over time make a mass market consumer version" (01:07).
So we're starting for something at the top end of the market, and you will see us over time make a mass market consumer version.
Source 3 · CNBC Television · 29 Sep 2026
OpenAI’s Sam Altman on new models: Safety has to stay ahead of capabilities
CNBC Television published a separate interview with him from DevDay on the same day.
THANK YOU SO MUCH FOR HAVING US HERE AT DEVELOPER DAY.
In the CNBC interview the host said "OpenAI essentially abandoned plans to launch a new model because of safety concerns" (00:37) and asked what had changed.
HAVE CONFIRMED IS THAT OPENAI ESSENTIALLY ABANDONED PLANS TO LAUNCH A NEW MODEL BECAUSE OF SAFETY CONCERNS.
Altman replied that "alignment, safety, monitoring, security have to stay way ahead of capabilities" (01:14).
ALIGNMENT, SAFETY, MONITORING, SECURITY HAVE TO STAY WAY AHEAD OF CAPABILITIES.
Show all 5 factsShow fewer facts
Asked what about the model worried him, he said "this model was a little bit worse on a few of the evals we look at" (01:35).
THIS MODEL WAS A LITTLE BIT WORSE ON A FEW OF THE EVALS WE LOOK AT.
He also said "think of this in the abundance of caution category" (01:40).
THINK OF THIS IN THE ABUNDANCE OF CAUTION CATEGORY.
Topics
The morning email
On the mornings we publish: the three top stories and up to four short ones. Free.
We email you a link to confirm. An issue may include one sponsor, always labelled Sponsored · Advertisement. Our emails count opens and clicks, not who made them. Unsubscribe in one click. What we keep
