House discloses China AI chip approvals and safeguards dispute | Magica
House releases China AI chip approvals, putting export safeguards to the test
Editorial Team
••📖6 min read
A bipartisan House disclosure shows requests to sell 3.4 million AI chips to China and approvals for more than 879,000. The oversight question is whether Commerce verified the buyers, supply protections and access restrictions behind those authorizations.
Requests covered 3.4 million Nvidia and AMD AI chips; approvals covered more than 879,000 as of May 2026.
Lawmakers flagged approved buyers designated as Chinese military companies and applications naming suspected smuggling fronts.
Approval totals establish neither deliveries nor compliance with restrictions on supply, military use and remote access.
The House Foreign Affairs Committee released export-license information on November 19 showing requests to sell 3.4 million advanced AI chips to China and approvals for more than 879,000. The committee said approved exports included buyers Alibaba and Tencent, which it identified as designated Chinese military companies.
The disclosure puts specific licensing decisions behind a dispute over the Trump administration’s conditional opening of China’s market to American AI chip suppliers Nvidia and Advanced Micro Devices (AMD). The Commerce Department’s Bureau of Industry and Security (BIS), which regulates and enforces exports, permitted sales subject to safeguards. Congress is questioning how the agency checked compliance.
Committee chairman Brian Mast, a Florida Republican, exercised disclosure authority under the Export Control Reform Act, the law underpinning BIS’s export-control powers. The committee cited a provision allowing release when withholding information would be contrary to the national interest. Members approved disclosure by unanimous consent.
House Foreign Affairs Committee figures show chips requested and authorized, not deliveries. The difference does not establish denials. Source: House Foreign Affairs Committee.
What the chip totals establish
The requests covered December through May, and approvals exceeded 879,000 chips as of May 2026, according to a report on the released documents. Requested buyers included ByteDance, Baidu and Tesla’s Shanghai operation.
Epic says it is using AI to rewrite parts of Unreal Engine and plans to let developers connect their preferred models to UE6. The strategy combines claimed maintenance gains with a broader overhaul of programming, distribution and portable game assets, while rival Unity already offers outside-model connections.
Editorial Team
These are quantities of chips requested and authorized, rather than numbers of license applications or a tally of chips delivered. Subtracting approvals from requests would not establish denials: the available accounts do not reconcile the balance by pending, withdrawn or rejected application. Nor do they provide transaction values or a breakdown sufficient to compare Nvidia’s and AMD’s shares of the business.
The buyers’ military-company designations make customer screening a central oversight question. But those designations alone do not establish which export restriction applied to each transaction or prove that a particular license was unlawful. That requires examining the parties and uses against the applicable licensing rules.
A conditional opening, with supply and access limits
BIS announced the policy change on January 13, following President Donald Trump’s December 8, 2025 announcement permitting Nvidia H200 and similar products to approved Chinese customers. H200 and AMD’s MI325X are advanced computing chips covered by the revised policy.
The rule took effect January 15, 2026. It replaced a presumption of denial with case-by-case review for qualifying exports from the United States to China and Macau. It did not grant unrestricted access to all advanced chips: eligibility depended on processing-performance and memory-bandwidth limits, and other applications remained subject to a presumption of denial.
That chronology matters. The disclosed request period began in December, before the rule took effect. The aggregate figures alone cannot show which policy governed each application or authorization.
To qualify for the new review policy, exporters had to certify, with supporting data, that exports would not delay U.S. customers’ orders beyond normal lead times or divert global chip-manufacturing capacity from similar or more advanced products for U.S. end users. BIS Under Secretary for Industry and Security Jeffrey Kessler defended controlled H200 sales in the January announcement as strengthening the American technology ecosystem.
The rule also required evidence that aggregate total processing performance exported to China or Macau would be no more than 50% of the corresponding performance shipped to customers in the United States for U.S. end use. The baseline runs from initial U.S. commercial shipments of the same advanced computing commodities through the license-application date. The disclosed chip counts cannot test that performance-based limit without product specifications, corresponding domestic shipments and application dates.
Customer controls extended beyond the buyer’s name. Applicants had to confirm the chips were not for prohibited military or military-intelligence users or uses, document customer-screening and physical-security procedures, and identify specified remote users. Facilities offering computing infrastructure as a service faced restrictions on remote access and transfers of model weights trained on the chips. BIS stated in the rule that it would routinely confirm the accuracy of relevant elements of applicants’ certifications.
Independent testing in the United States was another pre-export requirement. A qualified laboratory had to verify the shipment’s technical specifications and the exporter had to submit its certification to BIS before export; representative batch sampling was permitted. The rule allowed a commitment to submit the required certification before export, so a license approval alone does not establish that testing had already occurred.
Suspected intermediaries and changing routes to compute
Mast identified Maginfra and New Matrix, named in Nvidia applications, as suspected chip-smuggling fronts. The account of the meeting cited earlier New York Times reporting that the companies had empty offices and ties to companies under national-security sanctions. Mast framed authorization as an unanswered question: if BIS approved sales involving those firms, he said, that would be a serious concern.
Neither the applications nor that allegation establishes that BIS approved either intermediary. Maginfra and New Matrix did not immediately respond to requests for comment; Commerce, Nvidia and AMD also did not immediately respond to requests about the disclosure, the account said.
The release followed Jensen Huang’s resignation as Nvidia chief executive amid concerns about chips being diverted to China. Lawmakers from both parties have since called for stricter export controls. Commerce also issued an interim final rule that week targeting Chinese companies’ remote access to Nvidia computing power through overseas data centers, according to the same account.
Current supply is a separate question from the earlier approvals. Nvidia said it was “no longer shipping” chips including H200, according to a November 19 report. Alibaba, whose businesses include AI cloud and computing services, warned that the new U.S. rules could affect some global data-center expansion plans and said it was assessing the repercussions. Those company statements describe current constraints; they do not establish how many previously authorized chips were delivered.
The House Foreign Affairs Committee’s released table marks these entries “Approved, not shipped” as of May 11, 2026. It does not establish subsequent deliveries. Source: House Foreign Affairs Committee Democrats.
The records and testimony Congress needs next
Representative Gregory W. Meeks, the New York Democrat serving as the committee’s ranking member, said he had raised licensing questions for more than a year. In prepared remarks, he contrasted delays in BIS license reviews with what he described as insufficient scrutiny of advanced-chip approvals for China. His assertion that BIS failed to enforce the rules remains an oversight allegation.
Meeks called for public hearings with Commerce Secretary Howard Lutnick and Kessler. Huang was scheduled for a closed-door transcribed interview with the Foreign Affairs panel and two other House committees on December 1, followed by Nvidia general counsel Tim Teter on December 2, according to the report on the inquiry.
The next test is whether transaction records connect each authorization to the applicable policy, verified customers, domestic-supply evidence and pre-export certifications—and then to actual shipments. That record would let Congress distinguish a conditional sale properly authorized under the rules from a failure to enforce them.
At Ignite, Nvidia and Microsoft are expanding local coding AI, cloud desktop streaming and Azure infrastructure. The announcements build on an existing Windows AI strategy, with AMD alternatives, staggered availability and no comparable deployment-cost results in the announcements.
Microsoft and Nvidia are expanding Azure’s Vera Rubin virtual-machine preview, with systems already running in its data centers. Microsoft’s separate confidential-AI showcase covers four GPUs, while the announcements leave pricing, regions and general availability unspecified.
Meta is reconsidering its 2030 climate deadline after net emissions reached 11.8 million tonnes in 2025. Hardware and infrastructure purchases dominate its reported footprint, while new gas generation adds a longer-term challenge.
Three House committee chairs want the Remote Access Security Act in the annual defense bill as scrutiny of Nvidia intensifies. The available Senate text would authorize controls on specified foreign users’ remote access, with computing limits left to Commerce.
Google Quantum AI reports tests of quantum contextuality spanning up to 105 qubits on Willow processors. The experiments expose hardware errors and beat restricted classical strategies, while the largest task relies on an unproven classical benchmark.
OpenAI says its internal Performance Factory tracked more than 1,700 investigations in just over three months. About 400 cases reached proposed fixes, but far fewer shipped repairs used Codex’s code unchanged, and engineers retain approval.
Activists protested Anthropic’s stock-market debut outside Nasdaq, with at least one arrest and conflicting accounts of the total. Their campaign targets retirement-fund exposure to AI shares, alongside concerns about jobs, data centers and public accountability.
Anthropic says cash, growing revenue and multiyear payment schedules support its vast compute commitments. Its filing shows why flexibility matters: short-term capacity costs more, while major cloud contracts require payment regardless of usage.
OpenAI is rolling out live transcription for dictated ChatGPT prompts across free and paid plans. Users can see words as they speak and review the text before sending, but the company has not supplied performance measurements or a replacement recording maximum.
Boltz has raised a Bessemer-led $110 million Series A to expand molecular-design tools sold by usage. Its open-source research brought broad adoption; the commercial test is whether paid computation delivers repeatable gains in laboratory discovery.
Donald Trump said he and Xi Jinping supported minimal AI guardrails but had not discussed definitions. Their Shenzhen meeting leaves unresolved how bilateral safety cooperation would coexist with U.S. opposition to global regulation.
Gemini can now click and type across Mac apps and websites in a background window, Google says. The paid rollout expands its desktop automation, but app permissions expose active accounts and file changes that backups cannot always undo.
The House AI Safety and Innovation Caucus promises legislative advocacy and rapid incident briefings. Its members already have bipartisan safety proposals in play, making the choice of safeguards its next test.
Binance Wallet says it will return all subscription commitments, distribute roughly 2,957 Anthropic tokenized shares free and add a $1 million token giveaway after demand far exceeded its allocation.
Early Anthropic investor Dustin Moskovitz is urging frontier AI developers to slow down. He and Cari Tuna moved their stake to a nonprofit in early 2025, directing potential returns toward philanthropy while leaving questions about the holding's value and governance rights.
A national study finds higher odds of hosting data centers in majority-Black, Hispanic and Asian neighborhoods. Facility types, zoning rules and unequal access to public hearings complicate who bears the costs and receives the benefits.
A debut-day estimate values the Anthropic shares FTX sold in 2024 at $49.7 billion. The comparison exposes the scale of subsequent paper gains, but contrasts cash raised for bankruptcy repayments with a hypothetical stock-market valuation.
Anthropic raised $81.3 billion in a record IPO and opened 34% above its offer price before retreating. Rapid revenue growth and adjusted operating profits sit alongside vast computing commitments and limited investor control.
The Kimi developer reportedly abandoned planned Nvidia B300 capacity for 2027 after new US restrictions. The deal exposes financing risks for overseas AI clouds, but neither a model delay nor a $150 million cash loss is established.