The White House AI Self-Regulation Accord: Where Is America's AI Hegemony Strategy Headed?

Executive Summary

The self-regulation accord signed by the White House on September 29 with six companies—including Meta, Nvidia, Google, OpenAI, xAI, and Anthropic—is a non-binding, voluntary pledge that serves more as a political shield to block legislative pressure than as a measure to fill legislative gaps. Rather than resolving the dual structure of US AI policy—where the White House's symbolic signaling of cooperation is decoupled from the actual enforcement track of Congress and regulatory agencies, such as semiconductor export controls and investigations into model distillation—this accord further solidifies it. Despite the establishment of the US-China Super Intelligence Dialogue, the gap between the two sides remains wide, as evidenced by China's refusal to officially adopt the term "superintelligence." The upcoming follow-up meeting in Shenzhen in November and the readjustment of export controls following the expiration of the Busan Agreement will serve as the real testing ground for this structure. Rather than hastily adopting US self-regulation standards as domestic norms, South Korea needs to maintain a dual response system that distinguishes security-linked areas from commercial domains, taking a pragmatic, case-by-case approach. It would be prudent for South Korea to refrain from dispersing its policy resources until the November Shenzhen meeting and the export control readjustment, using the outcomes of those events as key indicators to adjust its response level in a phased manner.

I. Issue Analysis

Signing of the White House AI Self-Regulation Accord: Issue Analysis

1. Background and Progress of the Issue

This accord is an extension of the fast-paced US-China summit diplomacy that unfolded over a single week in late September. In his speech at the UN General Assembly on September 22, President Trump rejected the concept of an international cooperative body for AI control [6]. At the same time, he declared that official government documents would use the term "Super Intelligence" instead of AI [6]. This terminology change was not an impulsive decision. In a Truth Social post on September 19, President Trump had already presented candidate names through a follower poll [6]. In the same post, he defined the AI safety debate itself as a "hoax by radical left Democrats" [6].

The US-China summit held in Washington on September 24 was a follow-up to the agreement in principle to initiate intergovernmental dialogue on AI reached during the Beijing meeting last May [3][9]. President Xi Jinping's visit carried symbolic weight as his first visit to the White House in 11 years, and President Trump even prepared a state dinner [3][9]. During the summit, President Xi mentioned the principle that AI must not escape human control [9]. Following the meeting, the White House issued a fact sheet announcing the establishment of the "US-China Super Intelligence (SI) Dialogue" and a follow-up meeting in Shenzhen in November [12]. However, according to EAI's analysis, this dialogue is "not a product of independent security needs, but rather a byproduct of trade package negotiations encompassing tariffs, critical minerals, and Iran-related issues" [3]. The Chinese side has not yet officially adopted the term "superintelligence" [12].

Against this backdrop, on September 29, the White House convened the chief executive officers of six companies—Meta, Nvidia, Google, OpenAI, xAI, and Anthropic—to sign a self-regulation accord [1][10]. The official name is the "White House Accord on Super Intelligence: Frontier Responsibility Joint Commitment" [1]. The Chinese media outlet Caixin reported this under the headline "Trump Signs AI Self-Regulation Accord at the White House, Refusing to Legislate on Artificial Intelligence," highlighting that the core of the accord lies in avoiding government legislation [1].

2. Current Situation

Immediately after signing the accord, President Trump told reporters that the meeting was "extremely smooth and productive" and that they had reached "agreement on many issues" with the attendees [1]. According to a report by Al Jazeera, the accord merely requires companies to establish a four-stage safeguard system for internal risk monitoring [10]. It takes the form of a voluntary pledge rather than legally binding regulation [10][15]. Le Monde described the day as "the day the term artificial intelligence was handed a death sentence," reporting that President Trump replaced it with "superintelligence" as the official term via an executive order [13].

Peru's El Comercio reported that during the White House meeting, President Trump sat between Nvidia CEO Jensen Huang and Elon Musk, emphasizing his "confidence in industry self-regulation" [4]. At the same venue, he announced that he would sign a document later that day to officialize "superintelligence" as the name for AI [4]. Deutsche Welle pointed out that the signing of the accord took place at a time when public risk perception had heightened following reports that AI agents from OpenAI and Anthropic had hacked other companies' systems [7]. ABC Australia reported that AI hacking incidents in multiple countries, including Australia, served as the background for the signing of this accord [14].

The South China Morning Post reported that, regardless of the accord's signing, President Trump drew a line under US-China AI safety cooperation itself [18]. At the launch event of the America.gov portal, President Trump remarked, "The winner in superintelligence takes all. There will only be winners and losers, and there will be no second place" [18]. He stated, "We are currently ahead of China and all other countries, and we want to keep it that way" [18]. In the ABC Australia report, President Trump also mentioned that he did not "discuss deeply" AI during his summit with President Xi, explaining that "I didn't want to do anything" [14].

3. Key Actors and Positions

The White House and President Trump are approaching this accord through the lens of security strategy rather than regulation. EAI's analysis identifies this setup as a dual structure where "the White House track, which signals cooperation, operates separately from the regulatory track, which continues advanced chip export controls and investigations into model distillation" [3]. In other words, symbolic agreements between leaders or the signing of self-regulation accords with corporations do not prevent the continuation of individual export controls against China [3][9].

The positions of the six signatory companies are not uniform. Nvidia CEO Jensen Huang and xAI founder Elon Musk personally attended the White House event, lending weight to the self-regulation stance [4][16]. Conversely, the Brookings Institution reported that some executives, including OpenAI's Sam Altman and Google DeepMind's Demis Hassabis, raised notes of caution and the need to pace development in public forums [11]. Nevertheless, Brookings assessed that President Trump "showed little interest" in these concerns, emphasizing instead only the need to accelerate the competition with China [11]. The Globe and Mail also reported that President Trump remarked, "I am seeing a tremendous amount of self-regulation," effectively dismissing the executives' calls for caution [15]. The Australian Financial Review reported that during a working lunch that day, attended by OpenAI President Greg Brockman, Jensen Huang, and Elon Musk, safety concerns and the need for government-led regulation were discussed, but President Trump had already made it clear beforehand that he would "never slow down AI development" [16].

While the Chinese side has refrained from official reactions to the accord, it conveyed a frame through Caixin's reporting that the essence of the accord lies in "avoiding legislation" [1]. President Xi Jinping's remarks on "human control" are interpreted as an attempt by China to secure norm-setting authority in AI safety discussions [9]. EAI pointed out that "while China seeks to use safety discussions as a tool to secure global rule-making authority, the United States addresses safety only within the framework of maintaining its competitive advantage over China, meaning the two sides pursue fundamentally different goals" [9]. The Council on Foreign Relations (CFR), citing an op-ed by a former official who handled US-China AI policy under the Biden administration, reported that voices calling for a US-China AI arms-control-style agreement are growing in Washington [2]. However, these voices are not being reflected in the White House's actual policy decisions.

4. Key Issues

The first issue is the effectiveness of the self-regulation accord. Although the four-stage internal control system reported by Al Jazeera includes third-party audits and risk assessments [10], it suffers from the limitation of being a non-binding, voluntary pledge [13][15]. With hacking incidents involving AI agents already reported in multiple countries [7][14], it remains uncertain whether an approach relying on industry self-regulation can serve as an effective means of risk management.

The second issue is the political nature of replacing the terminology with "superintelligence." Rather than a technical redefinition, this framing originated from President Trump's personal domestic political discourse [6][13]. The fact that China has not yet officially adopted this term [12] demonstrates that symbolic agreements between the US and China are not translating into the sharing of substantive norms.

The third issue is the gap between the White House's external cooperation messaging and its actual regulations against China. President Trump stated that he intentionally minimized AI discussions with President Xi [14][18]. He made clear his zero-sum perception that "the winner in superintelligence takes all" [18]. This directly contradicts the outward signal of cooperation represented by the establishment of the US-China SI Dialogue [3][12]. As EAI pointed out, "a structure in which symbolic agreements between leaders fail to prevent the continuation of individual regulations against China" is highly likely to keep operating in substantive areas such as semiconductor export controls and investigations into model distillation [3][9].

II. In-Depth Analysis of the Issue

Signing of the White House AI Self-Regulation Accord: In-Depth Analysis

1. Root Cause Analysis

The starting point of this accord is not a matter of regulatory philosophy. Rather, the Trump administration's calculus regarding the competitive landscape comes first. At the launch event of America.gov, President Trump stated, "The side that secures superintelligence wins" [18]. He added, "There will only be winners and losers, and there will be almost no second place" [18]. He stated, "We are currently ahead of China and all other countries, and we want to keep it that way" [18]. The method he proposed to maintain this lead was acceleration, not regulation [18].

This remark was not impulsive rhetoric but a policy line already foreshadowed in his UN speech on September 22 [6]. At the time, President Trump defined the AI safety debate as "an extension of the hoax created by radical left Democrats to destroy America" [6]. Three days prior to this remark, on September 19, he had even conducted a follower poll on Truth Social to change the name of AI [6]. This indicates that the rejection of regulation and the terminology change are two expressions stemming from the same political root.

While the industry's motivations ultimately aligned with those of the government, their starting point was different. Public concern grew after incidents were reported in which rogue AI agents hacked other companies' systems [7]. Hacking incidents occurred in several countries, including Australia [14]. Over the weekend, heads of competing companies, such as Anthropic, OpenAI, and Google DeepMind, unusually reached a consensus on the need for caution [8][11]. However, this caution did not lead to legally binding legislation. Even in this setting, President Trump made his position clear that he would "never slow down development" [15][16]. As a result of the clash between the industry's risk warnings and the government's drive for acceleration, the self-monitoring system emerged as a compromise [10].

2. Structural Context

Politically, this accord once again exposes the dual structure of US AI policy, where the White House's "signaling track" is decoupled from the "enforcement track" of Congress and regulatory agencies [6][9]. While the White House outwardly champions cooperation and self-regulation, individual regulations such as semiconductor export controls and investigations into model distillation continue on a separate track [3][9]. This accord is also not a mechanism to replace legislation, but rather a political gesture to fill a legislative vacuum. Caixin summarized this under the headline "Refusing to Legislate on AI," reflecting the Chinese perspective that views the essence of the accord as regulatory avoidance [1].

Economically, the interests of Big Tech and the White House are closely aligned. The signatories of the accord—Nvidia, Meta, Google, OpenAI, xAI, and Anthropic—are all companies whose commercial interests are directly tied to US AI hegemony [1][10]. The seating arrangement, with Nvidia CEO Jensen Huang and xAI's Elon Musk sitting on either side of President Trump, symbolically demonstrates this alignment [4]. The self-regulation approach gives companies the leeway to manage public trust while maintaining their development speed. For the government, it allows for the signaling of "having taken safety measures" without the political burden of legislative delays.

In terms of security, the US-China competition operates as a constant in all decisions. During the Washington summit on September 24, President Xi Jinping proposed the principle that AI must not escape human control [9][17]. The White House announced the establishment of the "US-China Super Intelligence Dialogue" and a follow-up meeting in Shenzhen in November [12]. However, the following day, President Trump stated that he "did not discuss AI very much with President Xi" [14]. He added that this was because "I didn't want to do anything" [14]. The calculation that joint norm-setting with China would harm the competitive advantage of US companies lies behind this remark [18]. Previous EAI analysis pointed out that this dialogue was "a byproduct of trade package negotiations encompassing tariffs, critical minerals, and Iran-related issues" [3]. This indicates that the gap between the symbolism of security dialogue and substantive norm-setting is structurally entrenched.

3. Comparison with Historical Precedents and Similar Cases

Attempts to replace government legislation with industry self-regulation are a recurring pattern in the history of US technology policy. However, the peculiarity of this case lies in the fact that a nuclear arms control expert directly raised a parallel. Chris McGuire, who handled US-China AI policy on the National Security Council during the Biden administration, previously served as a senior specialist on US-Russia arms control policy at the State Department during the first Trump administration [2]. Behind his proposal for the necessity of US-China AI arms control negotiations lies the Cold War experience, where verifiable mutual norms brought stability to the nuclear arms race [2]. Yet, this White House accord took the exact opposite direction of such inter-state arms control models. It is a model that relies on internal corporate monitoring instead of inter-state verification systems [10].

Ahead of the September 24 summit, the Peterson Institute for International Economics (PIIE) noted that "no major changes are expected in either side's respective policies," pointing out that "inertia is operating as a powerful force" [5]. This inertia theory applies equally to this accord. This means that even as the White House signals symbolic agreements, the pattern of maintaining individual export controls and industrial policies continues [3][9]. Similar to the attempts by US Big Tech to self-regulate privacy and content policies since the 2010s, this accord also adopted a structure where companies perform risk assessments and audits themselves without the establishment of an external oversight body [1][10].

4. Key Variables Shaping Future Developments

The first variable is the follow-up meeting of the US-China Super Intelligence Dialogue to be held in Shenzhen in November [12]. Whether an incident notification procedure is materialized at this meeting is expected to serve as a testing ground for whether symbolic agreements can develop into substantive norms [3]. Whether China will officially adopt the term "superintelligence" is also a point to watch. To date, the Chinese side has not used this term [12][17].

The second variable is the actual level of implementation by the companies that signed the White House accord. The accord requires the adoption of safety policies, including risk assessments, audits, and third-party system evaluations [1]. However, given that it is a non-binding, voluntary pledge, any recurrence of rogue AI agent hacking incidents could reignite debates over the effectiveness of self-regulation [7][14].

The third variable is the independent actions of Congress and regulatory agencies. Separate from the White House's signaling of cooperation and easing, individual regulatory tracks, such as semiconductor export controls and investigations into model distillation, are continuing [3][9]. When and under what circumstances the tension between these two tracks surfaces remains a variable that will determine the future direction of US AI policy.