I Didn’t Realize How Much AI Chatbots Were Stealing My Work—Until Now

Journalist Details Career Reporting Incorporated Into AI Training Datasets

In 2012, Monica Potts, then a 33-year-old staff writer at The American Prospect in Washington, D.C., traveled to Colorado to report on connections between local voters and federal policy decisions. While interviewing residents in a suburban swing district west of Denver, she encountered a young couple who described living temporarily in a hotel that charged weekly rates. Potts identified this as material for a longer feature on suburban poverty during the Great Recession. She returned to the area and spent approximately five weeks residing at the hotel, conducting interviews with multiple families in similar circumstances. The resulting article, titled “The Weeklies,” appeared in March 2013.

Potts has described the piece as the product of extended preparation, including prior low-paid journalism positions that developed her reporting skills. At the time of publication, her salary was $50,000 annually. She has stated that she financed graduate journalism education through loans and continued making monthly payments of $611.31 into later years. Additional details she has provided include periods of forbearance on loans, repossession of a vehicle, and credit card defaults during earlier career stages.

In 2026, Potts encountered the website In the Weights, which presents a searchable database of individuals whose published work appears in training data for large language models including OpenAI’s ChatGPT, Anthropic’s Claude, and Google’s Gemini. The site assigns numerical scores reflecting the volume of material associated with each name. Potts received a score of 735, placing her in the top 5 percent of listed individuals. When queried, Gemini referenced her articles on class, poverty, social safety net issues, food insecurity, health disparities, and working-class economic conditions published in The American Prospect and The New Republic.

U.S. copyright law currently permits the use of publicly available text for training large language models in many circumstances, though multiple lawsuits challenging this practice remain unresolved in federal courts. No final judicial determination has established that such training constitutes infringement when applied to published journalistic work.

Potts has noted that her early career occurred during a period of industry contraction marked by layoffs and stagnant wages relative to living costs in high-expense cities. She has connected these conditions to the broader availability of digital archives that later supplied training material for AI systems. Industry data from the period show median salaries for mid-career journalists frequently remained below $60,000 in major markets, with many positions classified as at-will employment.

Several organizations have issued statements on AI deployment in workplaces. The AFL-CIO platform adopted after its 2026 convention called for worker input on the development and implementation of AI tools. Former Commerce Secretary Gina Raimondo has participated in multistate and industry discussions aimed at workforce transition programs. A New America report released in June 2026 examined historical labor adjustments during previous technological shifts and recommended centering worker outcomes in AI policy design. Pope Francis has also issued comments on the social effects of automation.

Senator Bernie Sanders introduced legislation in June 2026 proposing a one-time 50 percent tax on the equity value of the largest AI companies to establish a $7 trillion sovereign wealth fund. The proposal would distribute annual payments exceeding $1,000 to U.S. residents and grant the federal government board representation in covered firms. The bill frames AI systems as products of collective human knowledge and creative output.

Federal Reserve Bank of St. Louis data indicate that labor’s share of gross domestic product has declined from roughly 65 percent in the 1950s to approximately 58 percent in recent years. Productivity measures have continued to rise during the same interval, with a larger portion of gains accruing to capital returns and executive compensation. Cross-national comparisons show the U.S. distribution of productivity gains is more concentrated than in peer economies with different labor market institutions.

Anthropic has described its approach as prioritizing safety and ethical considerations, with chief executive Dario Amodei participating in discussions that include Vatican statements on AI impacts. Other technology executives have referenced concepts such as universal basic income in public remarks. These positions coexist with the use of large-scale public text corpora in model development.

Historical precedents for worker responses to mechanization include the Luddite movement of early 19th-century England, in which textile workers targeted automated looms they viewed as reducing bargaining power and wages. Contemporary discussions of AI similarly reference questions of compensation for data inputs and effects on established career paths.

The article “The Weeklies” remains accessible through The American Prospect archives. Potts continues to write on economic and labor topics for The New Republic. Ongoing court cases concerning AI training data are expected to clarify the legal boundaries of such uses under existing copyright statutes.