In mid-June 2026, within the same week, the Washington and state attorney general's offices almost simultaneously turned their guns towards two AI giants. On June 11, the U.S. government invoked national security authority to issue an export control directive to Anthropic, specifically calling for a suspension of all foreign nationals' access to the Fable 5 and Mythos 5 models; Anthropic, in order to ensure full compliance, simply expanded the scope to all users by directly shutting down access to these two models. Two days later, it was reported that multiple state attorneys general had jointly initiated an investigation into OpenAI, serving it with a sweeping subpoena the same week, demanding disclosure of advertising business, user activity and retention, consumer and health data handling, business information targeting minors and elderly populations, and even extending to the operation of deep learning models and internal company regulations. Thus, a top-down federal "hard line," justified in the name of export control and national security, and a bottom-up state-level "soft line," relying on consumer protection and competition laws, tightened simultaneously in the same time window, trapping Anthropic and OpenAI in the regulatory vise, while pushing an even more challenging question to the forefront: as compliance demands escalate, to what extent will AI companies be forced to compromise the technological boundaries and users' rights in the name of risk control.
Anthropic's blanket shutdown of Fable 5
On June 11, 2026, the export control directive from Washington was not broad in scope: the U.S. government used national security authority to require Anthropic to suspend access to the Fable 5 and Mythos 5 models for “all foreign nationals” (according to a single source). On paper, the regulation only drew a line based on nationality, isolating "foreigners" from the models; however, it was Anthropic itself that completely closed the door. Faced with a directive interpreted as triggering the national security clause of the Export Administration Regulations (EAR), this company, self-positioned as a cutting-edge AI safety firm, chose not to accurately identify user identities but instead announced, for the purpose of "ensuring complete compliance," to directly shut down access for all users— including U.S. citizens, foreign nationals, and expatriate employees—to Fable 5 and Mythos 5 (according to a single source). Following previous years of U.S. export controls on high-performance AI chips, this became the first explicit case extending to the access layer of AI models and services, and Anthropic, with a single "blanket" strategy, bound together technological boundaries, national security, and commercial services.
Once the axe fell, it was multinational users who first felt the disconnection. Regardless of where they were or what passports they held, as long as their workflows relied on Fable 5 or Mythos 5, they would soon face the reality of their access being entirely withdrawn: cross-border collaborative teams lost a unified model foundation, and globalized corporate clients were compelled to urgently adjust internal systems and compliance assessments. What was initially targeted at specific nationalities as export control, was amplified by Anthropic's decision into an immediate disruption affecting all users. For Anthropic itself, this "over-compliance" meant that in the short term, its core products would be taken offline, with potential revenue and customer trust under pressure simultaneously, but it also sent a clear signal to regulators: in an environment dominated by national security logic, a cutting-edge AI company is willing to sacrifice availability and business continuity in exchange for a policy safety net.
From chips to models: national security red lines tighten
From the timeline perspective, Anthropic merely stepped onto the extension of a line that had already been drawn. In the previous years, the U.S. had repeatedly ramped up export controls on high-performance AI chips, defining "computing power" itself as a strategic resource that could be severed; this time, the export control directive aimed at Fable 5 and Mythos 5 pushed the same logic to the model access level for the first time, making "what can be done" as important a review subject as "what is used to compute." According to a single source, these two models were considered to potentially trigger the national security clause of the U.S. Export Administration Regulations (EAR), meaning regulators legally recognized that even if no chip crosses borders, as long as high-capacity models are accessed by foreign nationals through the internet, they can be seen as a type of technology output that needs to be intercepted.
The national security clause is used here as a scalable ruler. It no longer only measures chip performance but begins to assess model capabilities and their potential spillover—what reasoning, generation, and decision-making functions will be regarded as impacting national security and thus fall within the scope of export control. The Anthropic incident thus became a clear precedent: once regulators confirm that a certain type of model is "sensitive enough," they can require companies to replicate this approach in access control, keeping specific nationalities and regions out of capability boundaries, leaving institutional space for other models or companies to be subjected to similar control in the future.
OpenAI under multi-state attack: advertising and data become the focal point
Almost in the same week that Anthropic was locked out of model access by national security clauses, another regulatory chain tightened at the local level. According to a single source, on June 13, 2026, multiple state attorneys general jointly launched an investigation into OpenAI and sent a broad subpoena that Friday, demanding the AI giant provide a large number of documents related to its operations and user impacts. This was not a simple information inquiry but a typical state-level law enforcement action: jointly initiated by elected attorneys general, symbolizing the first time "OpenAI" was formally examined within the coordinates of consumer protection and anti-unfair competition.
The subpoena outlined concerns that clearly delineated the risk landscape in the eyes of state regulators—on one end is the most commercially imaginative yet also the most sensitive advertising business, requiring explanations of how models push content and monetize user attention; on the other end are granular user behavior data: activity levels, retention curves, user dwell time and conversion in products, and how this data intertwines with consumer information and health data. More specific inquiries directly targeted minors and the elderly population, asking OpenAI to explain product design and protective measures for vulnerable users; meanwhile, the subpoena also included the deep learning models themselves, questioning the so-called model "flattery" issues and how company internal regulations constrain such behaviors. In the absence of a unified federal AI legislation, this joint investigation, leveraging consumer protection laws and anti-unfair competition laws, effectively signaled that even without a new specific AI law, state attorneys general could shape the behavioral boundaries of new technology using old tools, converting abstract "AI governance" into a series of concrete consumer cases.
Blocking while holding accountable: U.S. AI regulation dual tracks take shape
In the same week, Anthropic and OpenAI were each under scrutiny, but the logic was entirely different. On June 11, the federal government used national security authority to issue an export control directive, specifically requiring Anthropic to suspend all foreign nationals' access to the Fable 5 and Mythos 5 models; Anthropic, to ensure full compliance, directly cut off access for all users, including U.S. citizens and expatriate employees, effectively locking them out. This marked the first occurrence, following high-performance chip export controls, of the U.S. directly applying national security regulation to model and service access, targeting the risk of technology spillover beyond borders. Two days later, the OpenAI investigation, concentrated in media reporting, was spearheaded by multiple state attorneys general who, utilizing their consumer protection laws and anti-unfair competition regulations, pursued inquiries via subpoenas related to advertising practices, user activity and retention, consumer and health data handling, business design targeting minors and the elderly, and also included the deep learning models themselves, "flattery" behaviors, and internal company regulations, all packaged within the framework of "consumer rights."
The result is that, during a gap in the federal level lacking unified AI legislation, a federally driven export control line directed towards national security and a state-level investigation line oriented towards consumer protection and competition order both emerged almost concurrently in mid-June 2026, placing dual internal and external pressures on U.S. AI companies. Externally, companies must consider whether models can provide services to foreign nationals and which features could trigger national security clauses in export control regulations; internally, the operational details of the same model—what ads to push, how to retain users, how to handle health data, how to address minors and the elderly—might be interpreted at any time as infringing on consumers or distorting competition. As a result, regulatory attitudes showed a form of "internal and external difference": addressing cross-border outputs, discourse concentrated on technology spillover and national security; facing the domestic market emphasized corporate responsibility, user risks, and the acceptability of business models themselves. For companies, this means strategic planning must simultaneously encompass cross-border service frameworks and local compliance narratives, allocating costs for export compliance while preparing long-term investments to address state-level consumer inquiries, otherwise any product iteration risks being simultaneously halted on these two tracks.
How global AI companies can survive in the grey zone
Anthropic directly pulled the main power of Fable 5 and Mythos 5 under the export control directive, while OpenAI got dragged into the detailed mire of advertising, user behavior, and sensitive group data by state-level investigations. The picture painted by these two strands is that once large models provide cross-border services, they are no longer just "products," but are simultaneously placed under the scrutiny of national security and consumer protection. For global AI companies, the warning lies first in the "grey area cost"—in an environment lacking unified federal legislation and intertwining federal and state rules, capability boundaries, which users can access, and how data is used can all be reinterpreted at any time, making any iteration potentially forced toward extreme conservatism, similar to Anthropic. A more realistic survival strategy is to reserve a "safety margin" in model capability design, setting tiered access for users of different regions and identities, limiting the most sensitive capabilities to controllable groups; in data governance, establishing traceable and auditable internal standards related to advertisements, retention, and information for minors and the elderly, assuming "audits could come at any time"; on transparency, preparing an explanatory framework for regulators and the public in advance, clarifying what models can and cannot do, and how the company weighs benefits against risks, locking uncertainties within manageable ranges rather than betting that regulators won't come down hard.
Join our community to discuss and grow stronger together!
AiCoin exclusive Hyperliquid benefits: https://app.hyperliquid.xyz/join/AICOIN88
AiCoin exclusive Aster benefits: https://www.asterdex.com/zh-CN/referral/9C50e2
On-chain Telegram community: https://t.me/AiCoinWhaleData
On-chain community: https://www.aicoin.com/link/chat?cid=N6OVMor5g
AiCoin on-chain Twitter: https://x.com/aicoinwhaledata
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
