The prediction was that U.S. federal regulation would evolve from narrowly targeting AI to putting “basically all software companies” under direct federal regulatory oversight.

What has actually happened by November 2025:

• The most comprehensive federal move in 2023 was Biden’s Executive Order 14110 on AI, which focused on artificial intelligence systems (not the broader software industry) and emphasized frontier AI models and government use of AI. It did not create a general licensing or supervisory regime for all large software firms.(bidenwhitehouse.archives.gov) In January 2025, President Trump rescinded that order and replaced it with Executive Order 14179, which explicitly aims to remove barriers to AI innovation and directs agencies to revise or rescind prior AI‑safety measures viewed as restrictive. This is a move toward deregulation of AI, not expansion of oversight to all software companies.(en.wikipedia.org)

• Congress still has no broad AI or software‑sector regulatory statute. The only major new AI‑specific federal law is the 2025 TAKE IT DOWN Act, which narrowly targets non‑consensual intimate imagery and deepfakes and imposes takedown and process obligations on “covered platforms” that primarily host user‑generated content. It does not regulate the software industry as a whole or even all large software companies—only certain online platforms in a specific abuse context.(en.wikipedia.org)

• Other federal initiatives remain targeted and limited: export‑control adjustments for high‑end AI compute, voluntary NIST AI Risk Management guidance, and FTC enforcement of existing consumer‑protection law against misleading AI marketing claims. None of these create a new, comprehensive federal oversight regime that applies across “essentially all large software companies”; they are narrowly focused on AI risk or on specific content types.(en.wikipedia.org) Legislative proposals like the Preserving American Dominance in AI Act would focus on frontier AI and national‑security risks, not on the broader software industry, and as of late 2025 have not become law.(king.senate.gov)

• The regulatory trendline is also inconsistent with the prediction. Reporting on U.S. policy in 2025 describes federal moves to loosen or block regulation—such as attempts (ultimately defeated in the Senate) to impose a 10‑year federal moratorium on state AI laws via a tax‑and‑spending bill—rather than to build a strong, centralized federal regime overseeing all software firms.(reuters.com) At the same time, the most aggressive new AI rules have come from states like California (e.g., SB‑53, the Transparency in Frontier AI Act), which targets only very large frontier‑model developers and operates at the state level, not as a blanket federal framework for all large software companies.(en.wikipedia.org)

• Some separate, sectoral tech measures (such as the TikTok‑focused Protecting Americans from Foreign Adversary Controlled Applications Act) address specific national‑security concerns about foreign‑controlled apps, not general domestic software regulation.(en.wikipedia.org) Existing antitrust cases against firms like Apple are enforcement of long‑standing laws, not evidence of a new federal software‑industry regulatory overlay.(en.wikipedia.org)

Given this record two years after the prediction:

• Federal action has not broadened from AI to a regime that covers “basically all software companies” or “essentially all large software companies.”
• The only clearly new, binding AI‑specific federal statute is narrow (deepfakes/NCII and hosting platforms).
• The main comprehensive AI‑governance executive order was rolled back, and current federal policy direction is at least as much about removing restrictions and pre‑empting state efforts as about expanding federal oversight.

While “over time” is vague and does not specify an exact deadline, the prediction was directional (“we are headed to a place…”) and implied a trend toward blanket federal regulation of the software sector. The observable trajectory by late 2025 runs counter to that claim, so on the evidence available today, the prediction is best judged as wrong.