Marc Andreessen

Factual · AI, the economy and work

““because a significant percentage of jobs in the U.S. are licensed or unionized or civil service in a way where they literally cannot be replaced.””

Marc Andreessen, investor, podcast guest · 8 October 2025 · Billionaire investor Marc Andreessen says AI destroying jobs is a 'fallacy' (Fortune)

A significant percentage of U.S. jobs are licensed, unionized or civil service in a way that means they cannot literally be replaced.

Mixed Partly true and partly false in roughly equal measure.

The second judge would not resolve this claim; the verdict shown is the first judge's.

What would count as true or false

Judged as of 2025-10-08. The claim has two parts: (a) a significant share of U.S. jobs is licensed, unionized or civil service, and (b) these jobs 'literally cannot be replaced' (by AI or automation). True: authoritative data (BLS union membership, BLS/CPS occupational licensing estimates, government employment counts) show a substantial combined share, roughly 20%+ of U.S. jobs, AND evidence shows these protections legally or contractually bar those roles from being replaced. Mostly true: the share is substantial and the protections create strong legal or contractual barriers to replacement, so 'literally' is only a small overstatement. Mixed: the share is significant, but the protections only slow or complicate replacement rather than preventing it (for example, licensed tasks can be restructured, union contracts can be renegotiated, civil-service positions can be cut through layoffs or attrition). Mostly false: the share is real but the 'cannot be replaced' part is plainly contradicted by documented cases where such jobs were automated or eliminated, or the share is small (under ~10%). False: the data show the share is negligible, or that these categories give no meaningful protection against replacement.

Evidence

1. Union membership rates highest in Hawaii and New York, lowest in North Carolina in 2024 U.S. Bureau of Labor Statistics, The Economics Daily · 31 January 2025

   “The union membership rate in the United States (the percentage of wage and salary workers who were members of unions) was 9.9 percent in 2024, little changed from the prior year.”

2. Credentials U.S. Bureau of Labor Statistics, Occupational Requirements Survey · 1 January 2016

   “In 2025, 22.9 percent of civilian workers required a license or certification, 6.5 percent required an educational certificate, and 1.2 percent required an apprenticeship.”

3. Job security protections UC Berkeley Labor Center, Negotiating Tech inventory of U.S. union contract provisions · 15 July 2025

   “Generative AI shall not be used to enable the layoff of an employee or the elimination of a position covered by this Agreement or employee benefits under this Agreement.”

4. NLRA Protections for AI-Driven Layoffs? University of Chicago Law Review Online (Austin Smith) · 26 February 2026

   “Unions unable to secure these contractual protections have resorted to other measures.”

Reasoning

I rated this mixed: the evidence shows that a substantial share of U.S. jobs requires a license or certification or is unionized, but not that those jobs literally cannot be replaced. BLS data put union membership at 9.9 percent of wage and salary workers in 2024 (E1). Separately, 22.9 percent of civilian workers needed a license or certification in 2025 (E2). Together these meet the roughly 20%+ bar, though the E2 figure includes certifications as well as licenses. Some union contracts do bar using generative AI to cut jobs (E3), and a legal essay notes that unions unable to secure such contractual protections have resorted to other measures (E4). So these protections slow or complicate replacement in some cases rather than ruling it out.

Cite this claim

Marc Andreessen, “A significant percentage of U.S. jobs are licensed, unionized or civil service in a way that means they cannot literally be replaced.” (said 8 October 2025). Verdict: mixed. Reliability API, claim c_8cb09e732dd1, method 0.1.0+c876030c. https://epistacklabs.com/claims/c_8cb09e732dd1