Questioned about the specifics of those safeguards as he departed, President Trump informed reporters aboard Air Force One: 'standard safeguards that we discuss regularly.' Shipments of H200 chips to the ten approved Chinese purchasers are still delayed.

President Trump shared with reporters on Air Force One Friday that he and Xi Jinping addressed AI safeguards along with Nvidia's H200 chips over the two-day Beijing summit.

When pressed on the nature of those safeguards, the US president replied: 'standard safeguards that we discuss regularly.' He noted that the leaders had 'discussed potential collaboration' on the matter.

The summit concluded without any formalized AI governance agreement, and the key element under scrutiny failed to advance. Just prior to Thursday's talks, Washington approved about ten Chinese tech companies - including Alibaba, Tencent, ByteDance, JD.com, and Lenovo - to purchase up to 75,000 H200 chips apiece via a novel export-licensing system.

No H200 chips have been delivered to those approved buyers yet, while Chinese rare-earth exports linger roughly 50% below pre-control thresholds.

The term 'standard safeguards' carries significant weight in the summary, as the US and Chinese authorities have yet to publicly align on their scope.

Accounts of the meeting portrayed AI as 'the elephant in the room' rather than the focal point, with open discussions centering on trade and the H200 issue, while broader topics like autonomous weapons, AI model abuse, and dual-use applications were only sketched in general terms.

High-level officials speaking off the record indicated that both nations are exploring an ongoing dialogue mechanism for AI risks, though no timeline, task force, or documented commitment surfaced from these talks.

The export-licensing setup for those ten Chinese buyers is notably complex. H200 volumes destined for China are limited to no more than 50% of Nvidia's US domestic sales, every shipment requires validation by a US-based independent lab, buyers must vow non-military applications, and the arrangement mandates a 25% revenue share directed through US soil.

Thus far, this has resulted in approvals on paper without actual transfers of hardware.

Senate Democratic leader Chuck Schumer declared that 'granting China access to this top-tier US technology poses risks and endangers our dominance in the AI competition.'

The administration counters - as Nvidia CEO Jensen Huang emphasized last week - that the H200 lags one generation behind the Blackwell series, which stays under export restrictions, and channeling controlled sales to Chinese clients preserves revenue and American employment at home.

China's rare-earth restrictions, enacted last year in response to prior US tariffs, continue to disrupt Western supply chains for magnets and motors, and were not eased amid the AI exchanges. Both rare earths and semiconductors appear in the same bargaining category on official summaries from each side, yet saw no joint progress in Beijing.

The implications for industry are evident in the expanding AI investment wave. Major cloud providers have pledged over $650 billion for AI infrastructure through 2026, based on first-quarter reports from Microsoft, Alphabet, Amazon, Meta, and Apple, positioning Nvidia at the heart of that expenditure.

A revenue stream from China, even capped and with a 25% US conduit, significantly alters Nvidia's outlook. The Microsoft-OpenAI partnership represents the prominent US aspect; Huawei's Ascend processors embody the counterpart that the export rules aim to temper.

In this context, Trump's comments from Air Force One function more as a tactical update than a major AI policy reveal.

The president affirmed that AI safeguards and H200 chips featured prominently, with the week's licensing approvals still in effect for semiconductors. The anticipated formal document failed to materialize.