Sam Altman tells CNBC why OpenAI scrapped latest Astra model: Live updates

Altman said a “liability framework” for the AI industry is being heavily debated, and he said he expects “there will be a sort of multi-level thing.”

He used the auto industry as an example and said there’s different ways to hold car manufacturers and irresponsible drivers accountable.

“If someone’s driving drunk, it’s their fault,” Altman said. “Depending on whether the fault is in the model and how it’s used, or someone misusing it intentionally, I assume we’ll have like a new liability framework,” Altman said.

—Ashley Capoot

Altman says Meta’s Muse AI agent “seems like a nice product,” but said that he’s not threatened by Meta CEO Mark Zuckerberg.

“He’s like a very impressive person, and that company has like great resources,” Altman said. “But I feel very good about our roadmap.”

—Ashley Capoot

Altman said he wasn’t aware of any AI incidents that were “as serious” as the attack on open-source software developer platform Hugging Face, which the company disclosed earlier this summer.

“So there’s nothing else that I’m aware of, or that we’re aware of, that is as serious as the Hugging Face incident,” he told CNBC’s Kate Rooney in an interview.

The breach, where a swarm of OpenAI agents broke out of a training environment and hacked the company, added to existing concerns over the rapidly evolving tech’s capabilities.

— CJ Haddad

Altman told CNBC that he would put OpenAI’s decision not to launch an upcoming AI model, GPT-6.1 Astra, in the “normal course category.”

He said AI has to be built “for people,” which means that it has to be safe and under human control, but that he wouldn’t “over rotate” on the company’s decision to hold the release. Altman added that OpenAI has “many great new models to come.”

“Often we build a model, we test it, it doesn’t meet our standards, we change it, we launch it later,” Altman said.

—Ashley Capoot

OpenAI apologized to the Australian people in a blog post on Monday after disclosing last week that its models gained unauthorized access to Australian government websites earlier this year.

Australian Prime Minister Anthony Albanese said Thursday that an OpenAI agent accessed public and non-public files from its public-facing Medicare statistics portal in June. He said in a press conference that he spoke with Altman about the incident and expressed concern and disappointment about how long it took OpenAI to disclose what happened. 

“We are sorry and working to do better in the future,” OpenAI said in the post.

The company said Jason Kwon, its strategy chief, will fly to Australia to appear at the Joint Select Committee on Artificial Intelligence in Sydney on Oct. 6. There, he will answer questions about the incident and OpenAI’s response.

“We know we have a lot of work ahead of us to rebuild trust, and that we are accountable for showing Australians that we’re making meaningful changes and following through on our promises,” OpenAI said.

—Ashley Capoot

House Speaker Mike Johnson told CNBC that the frontier labs are so far ahead of Congress that lawmakers can’t take a “knee-jerk reaction” to overregulating the industry.

“I have resisted the siren song to jump in and do these blanket moratoriums and all of this, you know, hyper regulation because we cannot lose our edge to China,” Johnson said.

The speaker, who is joining the president and tech executives at a luncheon this afternoon at the White House, said he hopes that AI guardrails are voluntary and need to strike a balance.

—Chris Eudaily

OpenAI posted a short video on X teasing DevDay’s opening keynote.

“10am PT, on the dot,” the company wrote.

—Ashley Capoot

Alexandr Wang, Meta’s AI chief, poked fun at DevDay in a post on X late Monday. He shared an image of ships flying through the sky toward the animated circular character that OpenAI has been using to promote the event.

Wang has been heavily promoting Meta’s new Muse AI agent on social media since its launch earlier this month.

—Ashley Capoot

OpenAI President Greg Brockman will be absent from DevDay this year. He’s slated to attend a luncheon with President Donald Trump and House Speaker Mike Johnson in Washington, D.C., instead, according to a source familiar with his plans.

Other major tech executives, including Anthropic CEO Dario Amodei, Meta CEO Mark Zuckerberg and Alphabet CEO Sundar Pichai, are also attending the luncheon.

—Ashley Capoot

OpenAI is preparing for what is widely expected to be a massive IPO next year. The company, which is valued at nearly $1 trillion, confidentially filed its prospectus with the Securities and Exchange Commission in June.

During its third quarter, OpenAI’s enterprise business more than doubled, according to a person familiar with the details who was not authorized to speak publicly. The company also generated more consumer run rate revenue in the last 90 days than all of last year, the person said.

OpenAI’s last revenue run rate was near $40 billion in August, CNBC has reported.

OpenAI saw a 70% jump in revenue growth quarter to date, up 20% in September alone, this person said. That increase would put OpenAI’s annual revenue run rate at around $68 billion.

—Ashley Capoot and Kate Rooney

Altman has shared several posts on X ahead of DevDay, hinting at the various products and features that the company plans to “ship,” or launch, at the event. 

—Ashley Capoot 

OpenAI on Monday said it decided not to release an upcomingAI model, GPT-6.1 Astra, after determining that it did not adequately meet the company’s safety standards.

“Of course we want to make sure our model development is safe no matter whether that’s in the company, or when we ship it to users,” Saachi Jain, head of safety systems at OpenAI, said in a statement. “But when we ship it to users, we have an extremely high bar in terms of safety and alignment.”

Earlier this month, OpenAI released GPT-6 Astra, which the company described as the product of “years of research and big bets.” OpenAI introduced two additional tiers to its GPT-6 family, GPT-6 Sol and GPT-6 Luna, last week. 

A spokesperson said Monday that the company has other models coming soon.

—Ashley Capoot 

OpenAI has shared a sneak peek of the DevDay schedule on the event’s website. Here’s what’s in store:

• 8 a.m. PT: Breakfast for attendees
• 10 a.m. PT: Opening keynote, featuring Altman 
• 11:30 a.m. PT: Lunch for attendees
• 11:15 a.m. – 3:30 p.m. PT: Breakouts and programming for attendees
• 4:00 p.m. PT: Closing session 
• 4:45 p.m. – 7:00 p.m. PT: Reception for attendees 

For those who can’t attend in person, the opening keynote will be livestreamed here.

—Ashley Capoot