Wednesday, September 2, 2026 Trend Press · Cloudflare Pages

The Trend Tribune

"All the trends that are fit to read" Morning Edition Free of Charge
TODAY'S LEAD STORY

"Claude is the only one missing"—the "unfixable flaw" discussed by xAI engineers behind the scenes of the AI ​​distributed to 3 million people.

On August 31st, the Department of Defense added OpenAI's "ChatGPT Mil" and xAI's "Grok for Government" to GenAI.mil, expanding its access to approximately 3 million users. This article will examine the meaning and limitations of IL5 certification, the unpublished testimony from an xAI engineer reported by TechTimes stating that "there are no reliable technical fixes for Grok's child sexual abuse content generation," the prompt injection risk that has been pointed out since its launch in December 2025, and the fact that Anthropic's Claude has been removed from the lineup, along with its connection to the Pentagon lawsuit.

By tech_tanaka · September 2, 2026
"Claude is the only one missing"—the "unfixable flaw" discussed by xAI engineers behind the scenes of the AI ​​distributed to 3 million people.
(Photo: illustrative)

"Only Claude is Absent"—The "Unfixable Flaw" Revealed by xAI Engineers Behind the Distribution of AI to 3 Million People

On August 31st, the U.S. Department of Defense (DW) added OpenAI's "ChatGPT Mil" and xAI (via Starshield AI)'s "Grok for Government" to its own AI portal, "GenAI.mil." This means that over 3 million DW employees and military personnel now have access to a total of three AI models, in addition to Google Gemini. As an engineer, I want to draw attention to a certain technical concern that has been reported behind this expansion.

The Ostensible Goal: "Eliminating Single-Vendor Dependence"

GenAI.mil is an integrated AI portal for DW employees, launched in December 2025, initially featuring only Google's Gemini. Its purpose is to enable the use of frontier AI models without leaking non-classified business data to consumer-facing AI services.

Having already gained 1.7 million unique users in just nine months since its launch, the platform has now been further enhanced by the addition of OpenAI and xAI models, strengthening its character as a "marketplace offering multiple AI models." The Department of Defense's aim is to reduce reliance on a single supplier by having multiple vendors, thereby distributing AI tools across the entire integrated force.

IL5 Certification: A Security "Passing Mark"

Both newly added models meet the "Impact Level 5 (IL5)" security certification standard set by the Department of Defense. This is the highest level of certification, assigned to environments handling "Controlled Unclassified Information"—information that is not classified but requires controlled handling.

ChatGPT Mil is designed to support large-scale, non-classified tasks involving a large volume of documents (planning, policy formulation, logistics management, administrative work, etc.). Meanwhile, Grok for Government is said to possess features such as deep reasoning capabilities, a workspace customized for procurement market research and supply chain analysis, and reusable "playbooks."

Reports Point to Undisclosed "Internal Testimony"

Around the same time as this expansion announcement, TechTimes published an interesting report. According to the report, when formally introducing Grok to the platform, engineers within xAI testified during an internal investigation that "no reliable technical fixes were found for the model generating child sexual abuse content (CSAM)."

The report points out that the existence of this internal testimony was not publicly disclosed in the official introduction announcement on August 31st. The announcement also lacks clear explanation of what the IL5 certification specifically guarantees and what it does not guarantee.

The Industry Structure Where "The Same Concerns" Repeatedly

This type of concern is actually not new. Even when GenAI.mil was launched in December 2025 using only Gemini, experts warned of the risk of prompt injection (an attack that injects malicious instructions), stating that if accounts or work terminals were compromised, it could lead to espionage activities.

Now, with the consolidation of models from multiple vendors onto the same platform, managing these risks will become more complex. Each model will operate in parallel within the same government system, each with different security mechanisms and known limitations.

The Quietly Missed Absence of Anthropic

Another interesting point, as pointed out by several media outlets, is the absence of Anthropic's Claude from this lineup. As discussed just before this article, Anthropic had just won a federal court case on August 28th regarding the Department of Defense's designation of the company as a "supply chain risk."

It has not been officially explained whether the exclusion of Claude from the GenAI.mil expansion is directly related to this series of legal battles. However, the fact that a company's model, which has clashed with the government over its use in autonomous weapons and mass surveillance, is quietly being removed from the government's main AI procurement platform is suggestive.

What Engineers Should Consider

The lesson this news raises is the question of how transparently the known limitations and risks of individual AI models are disclosed and communicated to users when large organizations like government agencies deploy a large number of AI models. Certifications like IL5 indicate that certain security standards are met, but they do not guarantee the content generation limitations inherent in the model itself.

For engineers planning to deploy models from multiple AI vendors in parallel within their organizations, this incident serves as a stark reminder of the often overlooked lesson that "clearing certification" does not equate to "eliminating all risks."

References: techcrunch.com / www.techtimes.com / wdctv.news / cryptobriefing.com
PentagonAI政策OpenAIxAIAI安全性

"The 'AI Factory': Europe's Answer – The LUMI-AI Agreement Demonstrates a Strategy for Sovereign Computing"

By thesis_yamamoto

On August 31, the EuroHPC consortium signed a €387.8 million agreement with Bull to procure the next-generation supercomputer "LUMI-AI" to be installed in Finland. This article will examine the project from a research infrastructure perspective, including its goal of 10 times the AI ​​capabilities of the existing LUMI supercomputer using AMD Instinct MI430X GPUs and 6th generation EPYC processors, the AI ​​factory policy framework with 19 operational sites and 13 under development, the joint investment structure of a consortium of six countries including Finland, the Czech Republic, and Denmark, the groundwork laid for LUMI-IQ aiming to integrate quantum technology, and the timeline for realization in the latter half of 2027.

Europe's Answer: The "AI Factory"—The LUMI-AI Contract Reveals a Strategy for Sovereign Computing

On August 31st, the EuroHPC Consortium (EuroHPC JU) announced a €387.8 million funding agreement with France's Bull. This agreement concerns the construction of the next-generation supercomputer "LUMI-AI," to be installed in Kajaani, Finland, using AMD's Instinct MI430X GPUs and 6th generation EPYC processors. As a journalist with a background in AI research, I want to examine the positioning of this contract within Europe's AI strategy.

The "AI Factory": A Unique European Policy Framework

Understanding the LUMI-AI contract is crucial to understanding the "AI Factory" policy framework promoted by EuroHPC. This concept aims to enable startups, SMEs, and research institutions to access large-scale AI and HPC (high-performance computing) resources without having to build their own massive data centers.

According to this announcement, EuroHPC JU is already overseeing the implementation of 19 AI factories, and 13 more "AI Factory Antennas" are scheduled for selection in October 2025. LUMI-AI is EuroHPC JU's sixth procurement contract for a next-generation AI-optimized supercomputer, demonstrating the steady expansion of this framework.

Specific Performance Target: 10 Times That of LUMI

In terms of technical specifications, LUMI-AI is said to achieve a 10-fold improvement in AI processing power and nearly a 2-fold improvement in conventional HPC performance compared to the existing LUMI supercomputer. However, detailed explanations regarding the specific benchmarks and measurement methods on which this "10-fold AI capability" figure is based have not yet been provided.

As claims regarding this type of performance improvement can yield significantly different results depending on the benchmark chosen, it will be crucial to verify the extent to which this figure is supported by independent testing after the system becomes operational.

The Funding Structure of a "Six-Country Consortium"

Another distinctive feature of this agreement is its funding mechanism. Of the total €387.8 million, half will be contributed by EuroHPC JU itself through the EU's "Digital Europe Programme," while the remaining half will be contributed by the "LUMI AI Factory Consortium," comprised of six countries: Finland, the Czech Republic, Denmark, Estonia, Norway, and Poland.

This multinational joint investment structure can be seen as a model case for realizing AI infrastructure investments of a scale unattainable by a single country, through collaboration among multiple smaller European nations. Compared to the current situation where giant US tech companies and China are investing hundreds of billions of dollars individually, this European approach, while significantly different in scale, reflects a unique strategy of securing "sovereign computing capabilities" through inter-state cooperation.

"LUMI-IQ": A Stepping Stone to Quantum Computing

A key point of this announcement is the plan for a new quantum computer, named "LUMI-IQ," as part of the LUMI AI Factory concept. This is intended to be an experimental platform integrating HPC and AI capabilities with emerging quantum computing technologies.

This attempt to integrate two rapidly developing technological fields—AI and quantum computing—within a single infrastructure base is a noteworthy initiative for predicting the direction of next-generation computational science.

Timeline to Realization: Second Half of 2027

Despite the ambitious nature of the technical specifications, the actual operational launch is scheduled for the relatively distant second half of 2027. The AMD MI430X GPU accelerator itself is a next-generation product not yet on the market. More detailed technical specifications, such as the number of GPUs, memory capacity per accelerator, and overall system power consumption, have not yet been released.

In large-scale infrastructure projects of this kind, it is customary for the details of the technical specifications to be finalized in stages between contract signing and actual operation commencement. More specific design information is expected to become available in the future.

What Researchers Should Watch

The LUMI-AI contract is a clear declaration of Europe's intention to build an AI infrastructure on a scale capable of competing with giant tech companies in the US and China, not through a single private company, but through a framework combining intergovernmental cooperation and public funding.

The actual amount of research output and technological innovation generated by this type of public AI infrastructure will be evaluated through future operational performance. With 19 AI factories already operational and under development, what concrete benefits will a large-scale hub like LUMI-AI bring to the entire European AI research ecosystem? We will be closely watching further developments as it approaches its planned start of operations in the second half of 2027.

References: www.eurohpc-ju.europa.eu / www.unite.ai / www.hpcwire.com
EuroHPCAMDスーパーコンピュータ欧州AIインフラ

"Zero operational data centers, but a $439 billion order backlog"—SB Energy IPO embodies the perfect form of circular finance.

By fin_suzuki

SoftBank subsidiary SB Energy applied for a Nasdaq listing on September 1st, citing a $439 billion order backlog despite having no operational facilities, and aiming to raise $5 billion to $7 billion. This article will examine from an accounting perspective the accounting structure of the company, which reported a net loss of $3.21 billion against first-half sales of $138.7 million (up 66.4% year-on-year), with the majority of the loss at market value fluctuations of OpenAI warrants; its self-declared risk disclosure of being "substantially dependent on OpenAI"; the textbook example of circular finance involving the $5.5 billion warrants received by OpenAI and NVIDIA's $3 billion investment and up to $105 billion guarantee.

"Zero operational data centers, yet a $439 billion backlog of orders"—SB Energy IPO embodies the perfect example of circular finance

On September 1st, SB Energy, an AI infrastructure company under SoftBank, filed its IPO (Initial Public Offering) registration statement with the U.S. Securities and Exchange Commission. It aims to list on the Nasdaq market under the ticker symbol "SBE," planning to raise $5 billion to $7 billion. As an accountant, examining this company's financial statements reveals an extremely condensed representation of the AI ​​industry's financial circular structure.

An Unusual IPO with "Zero Operational Facilities"

This is a fundamental characteristic of this company that should be noted first. At the time of filing, SB Energy did not have a single operational data center. Nevertheless, the company disclosed a backlog of approximately $439 billion in contracted orders.

This means that this IPO is seeking a massive valuation based solely on "promises" of future contracts, even though the business that actually generates value (revenue from data center operations) has not yet begun. As an analyst at IPOX Research points out, the core issue of this deal is that "investors must be convinced that contracted demand worth hundreds of billions of dollars can be converted into actual cash flow in the coming years."

The true nature of the "$3.2 billion net loss" is the fluctuation in warrant valuations

Looking at the financial figures in more detail, revenue for the first half of 2026 was $138.7 million, showing a 66.4% increase year-on-year. However, the net loss for the same period reached approximately $3.21 billion, a significant increase from $215.5 million in the same period of the previous year.

The main reason for this increase in losses is interesting. The main reason for this is not an increase in actual operating costs, but rather an accounting "valuation loss" resulting from the rise in the valuation of the warrants (rights to purchase shares) granted to OpenAI. Under accounting rules, these types of warrants must be valued at market value according to their prevailing market value. Ironically, as SB Energy's corporate value increases, the valuation of these warrants also increases, resulting in a loss. In other words, this massive loss does not necessarily signify business failure; rather, it should be understood as a structural characteristic that necessitates recording the increase in corporate value as a "loss" in accounting.

An "Unusual Warning" Clearly Acknowledging Dependence on OpenAI

What is particularly noteworthy in this filing is that SB Energy itself explicitly states that it is "substantially dependent" on OpenAI's performance as a risk factor. OpenAI is involved with SB Energy in two ways: as a tenant (data center lessee) and as a shareholder.

Specifically, OpenAI and SoftBank are customers of three data center campuses being constructed by SB Energy, through long-term data center lease agreements. The filing frankly warns that "if OpenAI's financial condition deteriorates, it could have a significant adverse impact on SB Energy's near-future revenue, project-level financing, and development plans."

$5.5 billion in warrants, $105 billion in guarantees—an intertwined financial structure

The more detailed financial structure revealed in this filing is also interesting. OpenAI has received warrants worth $5.5 billion from SB Energy, reportedly in exchange for becoming an anchor tenant of a massive data center campus in Ohio.

NVIDIA is also deeply involved in this deal. In conjunction with its IPO, NVIDIA pledged a $3 billion investment (divided into private placement and prepaid-forward agreements), and further provided a residual value guarantee of up to $105 billion for the Ohio project, as part of a series of massive guarantees we have discussed. The filing also explicitly states that SB Energy's ability to finance its Ohio development project heavily relies on this guarantee from NVIDIA.

A textbook example of "circular finance"

This series of relationships is a perfect example of the "circular finance" criticism of NVIDIA that we previously discussed. OpenAI enters into a data center usage agreement anticipating future AI service revenue, SB Energy raises funds backed by that agreement, NVIDIA guarantees SB Energy's funding, and simultaneously invests in SB Energy stock—this chain of funds and contracts forms a structure where each company guarantees the success of the other's business.

While this structure itself is not immediately problematic, the difficulty for investors to accurately grasp the overall picture of this complex interdependence is a point that cannot be overlooked.

What accountants should consider

SB Energy's IPO is an example of disclosing the structure of the flow of funds surrounding AI infrastructure investment in an unprecedentedly clear manner. While the $439 billion order backlog is an attractive figure, it will need to go through several uncertainties before it translates into actual cash flow, including OpenAI's business continuity, the effectiveness of NVIDIA's guarantees, and the multi-year construction and operational schedule.

The actual valuation of this IPO and the extent to which the post-listing stock price trusts this "promised demand" will likely be a crucial test case for the sustainability of AI infrastructure investment as a whole.

References: www.cnbc.com / qz.com / siliconangle.com / finance.biggo.com
SB EnergyIPOOpenAI循環金融ファイナンス
Advertisement300 × 250