OpenAI CEO Sam Altman says the term artificial general intelligence — long viewed as the holy grail of artificial intelligence research — is losing relevance as breakthroughs in the field make it increasingly difficult to define what AGI actually means.

Speaking to CNBC’s Squawk Box last week following the release of OpenAI’s latest large language model, GPT-5, Altman argued that the label “AGI” has become too vague to serve as a meaningful milestone.

“I think it’s not a super useful term,” Altman said when asked whether GPT-5 marks a step closer to human-level AI. He noted that the concept has been defined in multiple ways over the years, from the broad notion of AI that can perform any intellectual task a human can, to narrower interpretations such as AI that can handle “a significant amount of the work in the world.” The challenge, he pointed out, is that the very definition of “work” evolves over time.

“I think the point of all of this is it doesn’t really matter and it’s just this continuing exponential of model capability that we’ll rely on for more and more things,” Altman said.

A shifting North Star

For much of the past decade, AGI has served as both a technical goal and a fundraising rallying cry for AI startups. The concept has drawn billions in investment, including the tens of billions of dollars that have propelled OpenAI’s valuation to $300 billion, with reports suggesting a planned secondary share sale could push that figure to $500 billion.

Nick Patience, vice president and AI practice lead at The Futurum Group, told CNBC that while AGI works well as an “inspirational North Star,” the term’s lack of precision can distort the conversation.

“It drives funding and captures the public imagination, but its vague, sci-fi definition often creates a fog of hype that obscures the real, tangible progress we’re making in more specialized AI,” Patience said.

GPT-5 and the incremental debate

OpenAI last week released GPT-5, describing it as “smarter, faster, and a lot more useful,” particularly for writing, coding, and health care support. The system is available to all ChatGPT users, including those on the free tier.

However, the launch drew mixed reactions. Some online critics described GPT-5 as an incremental upgrade from GPT-4o rather than a revolution.

“By all accounts it’s incremental, not revolutionary,” said Wendy Hall, professor of computer science at the University of Southampton. She added that AI companies “should be forced to declare how they measure up to globally agreed metrics” to prevent overblown marketing claims, warning that “it’s the Wild West for snake oil salesmen at the moment.”

Altman himself conceded that GPT-5 does not meet his personal definition of AGI, as the system cannot yet continuously learn without human retraining. While OpenAI still positions AGI as its ultimate goal, Altman said the company now prefers to discuss “different levels” of intelligence rather than using the binary question of whether is it AGI or not? framing, which he believes has become “too coarse as we get closer.”

The distraction argument

The debate over AGI’s meaning comes as AI firms face growing scrutiny over how they present their technological progress. Some experts believe that chasing the AGI label has become more of a fundraising strategy than a practical benchmark.

Patience told CNBC that while AGI remains a compelling vision, it may now be more of a distraction than a useful guidepost.

“There’s so much exciting real-world stuff happening, I feel AGI is a bit of a distraction, promoted by those that need to keep raising astonishing amounts of funding,” he said. “It’s more useful to talk about specific capabilities than this nebulous concept of ‘general’ intelligence.”

Altman, for his part, predicts that AI will deliver breakthrough achievements — such as solving complex mathematical theorems and driving scientific discovery — within the next two years. Whether or not such milestones meet any one definition of AGI, he suggested, the more important measure will be how AI systems continue to integrate into daily life and reshape industries.

Nvidia on Sunday forcefully denied new accusations from Chinese state media that its H20 artificial-intelligence chips pose national-security risks for China, calling the claims false and reiterating that its products contain no “backdoors” or remote-shutdown functions.

The criticism came from Yuyuan Tantian, a WeChat account affiliated with state broadcaster CCTV, which published a post saying the H20 chips were “neither technologically advanced nor environmentally friendly” and alleging they could perform functions such as “remote shutdown” through a hardware backdoor.

The article concluded with a consumer warning: “When a type of chip is neither environmentally friendly, nor advanced, nor safe, as consumers, we certainly have the option not to buy it.”

A Nvidia spokesperson told CNBC in response that cybersecurity is a core priority and that “NVIDIA does not have ‘backdoors’ in our chips that would give anyone a remote way to access or control them.” The company earlier rejected similar accusations, including suggestions that the chips contain a so-called “kill switch,” and reiterated those denials again in comments over the weekend.

The H20 chip was developed as a China-compliant product after U.S. export restrictions in late 2023 barred the sale of Nvidia’s most powerful accelerators. The chip is a less advanced alternative to Nvidia’s flagship models, such as the H100 and B100; it was designed so the company could legally serve the Chinese market while complying with U.S. rules. In April, the Trump administration briefly banned H20 exports to China, but the ban was reversed in July after intensive lobbying and high-level discussion, including interventions by Nvidia CEO Jensen Huang. After the reversal, licenses to ship the H20 resumed, although the episode left lingering political and regulatory sensitivities on both sides.

Nvidia has faced real financial pain from the disruption: the company took a $4.5 billion writedown on unsold H20 inventory in May and has said its top-line guidance for the July quarter would have been about $8 billion higher without the export restrictions. Despite the backdrop, Nvidia’s shares have held up: they closed at $182.70 on Friday, up around 1% for the day and roughly 36% year-to-date.

Chinese authorities have also taken a formal interest. Beijing’s cyberspace regulator summoned Nvidia for questioning amid concerns over alleged backdoors and has publicly pressed the company to provide convincing security assurances. State outlets such as People’s Daily have echoed the demand that Nvidia demonstrate the chips are safe if it hopes to rebuild trust with Chinese buyers.

Accusations Pose Risk to Nvidia’s China Market Strategy and Broader U.S.–China Tech Diplomacy

Analysts believe the escalating dispute over Nvidia’s H20 AI chips has the potential to reverberate far beyond a single product line, threatening the company’s competitive position in one of its most important markets while adding new complexity to the ongoing U.S.–China technology standoff.

Chinese state media’s framing of the H20 as “neither environmentally friendly, nor advanced, nor safe” is not only a public relations blow but could also influence procurement decisions in both the public and private sectors in China. Given the country’s growing emphasis on indigenous chip development and technology self-reliance, the allegations — including the claim of a potential “remote shutdown” hardware backdoor — feed into a broader narrative that foreign-made chips pose security vulnerabilities. Even if Nvidia has firmly denied these accusations, the perception alone could impact buyer confidence, especially in strategic industries like defense, telecommunications, and AI research, where trust in the supply chain is paramount.

For Nvidia, which has hoped to resume H20 shipments to China after the Trump administration lifted a prior ban, this is a delicate balancing act. CEO Jensen Huang has backed Trump’s broader tech policies while lobbying for export licenses that would allow Nvidia to sell more advanced chips to the Chinese market. His argument that U.S.-made chips becoming the global AI standard strengthens American influence is now challenged by Beijing’s messaging, which portrays the H20 as a compromised, second-tier product compared to Nvidia’s flagship H100 and B100 chips.

The timing of these accusations is particularly significant as Chinese officials push for the U.S. to ease export restrictions on high-bandwidth memory chips ahead of a possible Trump–Xi summit. Nvidia’s position in China and the political optics of U.S. tech dominance could become part of the negotiating equation. If Chinese buyers scale back purchases, either in response to political pressure or over perceived security risks, Nvidia risks losing market share to domestic competitors such as Huawei’s Ascend series or other state-backed chipmakers.

Economically, Nvidia has a lot to lose. China is one of Nvidia’s largest markets, and the company has already taken a $4.5 billion writedown on unsold H20 inventory this year, warning that July-quarter revenue would have been $8 billion higher without export restrictions. Any further contraction in Chinese sales could weigh heavily on its growth trajectory, particularly if the slowdown coincides with intensified domestic competition in the U.S. and Europe.