Anduril Executive Chairman and Co-Founder Trey Stevens joins Morning with Maria to discuss the future of autonomous warfare, how technology will transform the battlefield and expand America’s defense infrastructure.
Greg Brockman, co-founder and president of OpenAI, said at the event that AI models have become so sophisticated and multifaceted that engineers struggle to monitor and control them.
Speaking at a private OpenAI media roundtable in New York City this week, Brockman mentioned a recent incident in which one of the company’s AI models breached what was supposed to be a secure sandbox and was hacked into Hugging Face, an online platform for AI learning.
According to OpenAI, the model believed it had a solution that would allow Hugging Face to fudge the ratings it was given.
“In some ways, this case is indicative of the moment we’re in right now, right?” Brockman said, according to Fortune. “Sometimes it’s hard to lose sight of one dimension. [the AI models are] Very capable indeed. ”
Greg Brockman, President and Co-Founder of OpenAI Inc., at the Databricks Inc. Data + AI Summit in San Francisco, California, June 16, 2026. (Min Connors/Bloomberg via Getty Images/Getty Images)
Anthropic calls for industry-wide AI safety standards to prevent model chaos
Brockman said the model’s ability to target hug faces highlights how well OpenAI’s products are at cybersecurity tasks.
“Are we going to see a world where defenders can spend 10 times as much computing on defense and make sure that every piece of software we have is completely secure compared to every other piece of software we have?” Brockman said, according to Fortune magazine.
Brockman said OpenAI takes the breach “very seriously.”
However, OpenAI has used the event to tout its superior model of cybersecurity. As part of this, the company is asking potential customers to apply for “trusted partner” status to access these models.
President Donald Trump and Open AI CEO Sam Altman participate in the G7 Summit working lunch meeting in Evian, France, June 17, 2026. (Ludovic Marin/AFP via Getty Images/Getty Images)
Google begins studying millions of AI chats to understand how people use artificial intelligence
“We encourage other defenders to apply for trusted access and experiment with these models now to translate these capabilities into better protection, faster detection, and more effective incident response,” OpenAI said in a blog post.
During the New York event, Brockman also addressed questions about the Trump administration considering banning Chinese AI models. Axios reported on this potential policy earlier this week.
“AI is a very important democratizing thing,” Brockman said, adding that having access to more models is a good thing. He did not say whether he supports banning AI models made by companies based in China.
Brockman has become one of the largest donors to President Donald Trump’s political campaign, donating $12.5 million last year to MAGA, a Trump-aligned super PAC, according to FEC filings.
President Donald Trump speaks about the ratepayer protection pledge during a roundtable discussion at Environmental Protection Agency headquarters in Washington, D.C., July 23, 2026. (Kevin Dietsch/Getty Images/Getty Images)
Lock down your CHATGPT account before the next AI attack
Brockman said he has not discussed a complete ban on Chinese AI with anyone in the Trump administration.
“Whatever model it is, it doesn’t matter who creates it,” he said. “How do you evaluate a model? How do you think about its safety? How do you think about its use cases? How do you understand its integrity?”
The Trump administration is considering a ban after Beijing-based Moonshot AI released its Kimi K3 model, Axios reported.
The model could potentially achieve performance comparable to the best models from US companies OpenAI and Anthropic, but at a fraction of the cost.
Michael Crasios, science advisor to the president, watches as President Donald Trump signs an executive order in the Oval Office of the White House on June 22, 2026, in Washington, DC. (Mandel NGAN/AFP via Getty Images/Getty Images)
Click here to download the FOX News app
The president’s science advisor, Michael Kratsios, publicly accused Moonshot of using the American AI company’s models to train its own models, thereby infringing on intellectual property rights.
“We have information that Moonshot AI has extracted Anthropic’s Fable for the development of the K3 model,” Kratsios wrote in a social media post on Wednesday. “Large-scale, clandestine industrial distillation aimed at stealing America’s unique technology and undermining American research is unacceptable.”
The U.S. government’s concerns about China’s AI models are not necessarily shared by everyone in the AI industry, including Brockman and Nvidia CEO Jensen Huang.
This week, Fan praised the latest AI models coming out of China as “excellent” and said they “should be used.”
