On September 29, 2026, President Trump signed Executive Order 14434, "Inaugurating the Era of Super Intelligence," directing every executive-branch agency to replace "Artificial Intelligence" and "AI" with "Super Intelligence" and "SI" — and in the same afternoon, OpenAI, Anthropic, Google, Meta, xAI, and Nvidia signed a voluntary safety accord committing their frontier models to four layers of internal and external oversight.
The order is a terminology mandate, not a hard regulatory regime. According to the White House fact sheet, it "directs all departments and agencies to use the terms 'Super Intelligence' and 'SI' in official correspondence, public communications, policy documents, and non-statutory documents within the executive branch," and agencies are told to "no longer acknowledge" the older terms. Statutory language and independent federal bodies are outside its reach, so the practical effect is on federal prose, not on the law.
The accord's four layers
The more consequential piece is the one-page industry pledge. CBS News reported that the tech companies "committed to 'four layers of controls and audits' to ensure the AI models are acting as intended, including internal evaluations, audits by an external firm and reviews by each company's board of directors." The document, named the "Joint Commitment on Frontier Responsibilities," was signed "after a White House lunch with more than 20 AI and tech executives," according to IJR.
- Layer one: internal controls monitoring model capabilities and alignment during training and deployment, covering cybersecurity, biosecurity, and chemical threats.
- Layer two: a dedicated internal team ensuring those controls, monitoring, and detection systems operate as intended.
- Layer three: independent external auditors evaluating whether safeguards actually work.
- Layer four: independent board committees overseeing the process and reviewing reports from internal and external reviewers.
The six signatories sent an unmatched concentration of AI decision-makers: Google CEO Sundar Pichai, Anthropic CEO Dario Amodei, Meta CEO Mark Zuckerberg, OpenAI President Greg Brockman, xAI CEO Elon Musk, and Nvidia CEO Jensen Huang, per The Economic Times. Trump called the pledge "morally binding," according to TechCrunch's recap of the White House gathering. The Register was blunter, greeting the signing with the headline "weak, non-binding, AI regulations." There is no enforcement mechanism and no penalty for noncompliance written into the accord.
What this means for companion and creative AI
These six companies are the gatekeepers of the model APIs that power most of the companion market, and every one of them has spent 2026 defending or tightening content restrictions that define what companion and NSFW AI tools can do. Anthropic keeps Claude's adult-roleplay and explicit creative-writing blocks in place; OpenAI prohibits sexual roleplay in ChatGPT and just shipped teen-specific restrictions; Meta's Muse ships with a hard NSFW block; xAI pulled Grok's companions into the separate Animates app after removing them from the main product.
An external-audit and board-oversight regime at exactly those companies could push content policy in either direction: tighter filters justified by audit findings, or more documented, defensible release criteria. But the accord's four layers are all aimed at national-security-style model risks — cyber, bio, chemical — not at the consumer-facing problems that define the companion beat: emotional dependency, age verification, deceptive personas, or the NSFW policy fights that drove a wave of state laws from California SB 243 to Tennessee's companion statute.
Bottom line
The rebrand is noise. The structure is the signal: the White House blessed an oversight architecture at the companies whose policies control the companion market, while leaving it voluntary, omitting consumer protections, and pushing companion-specific regulation down to the states. The first independent audit reports will show whether "Super Intelligence" oversight is a real constraint or a photo-op — and companion users should read them for what they reveal about the content filters their apps run on.