Did AI Draft Trump’s Tariff Formula? A Deep Dive into ChatGPT’s Trade Model

Media outlets reported that Trump's April 2 tariff policy might have been generated by AI.

After entering the question into ChatGPT, the author received a model that closely matches the official formula.

ChatGPT’s “reverse‑deficit tax” model

✅ Basic principle: impose tariffs on U.S. exports to each country to punish long‑standing trade deficits and protect domestic manufacturing.

📊 Model suggestion: dynamic tax based on the trade‑deficit ratio.

Step 1: Define core metric

Define “deficit ratio” as the amount the U.S. imports from a country minus what it exports, divided by total imports – the higher the ratio, the larger the tariff.

Step 2: Set basic tax logic

The minimum rate is set to 10 %. A simple linear model can be used:

t_i = r_i × k, where t_i is the tariff rate for country i, r_i is its deficit ratio, and k is a scaling factor (e.g., 50 %).

Example: if a country’s deficit ratio is X and k = 0.5, the resulting tariff is …

🔁 Optional advanced scheme (policy smoothing): use an average deficit over several years and cap annual rate increases (e.g., no more than +5 %).

🧮 Summary of steps:

Obtain the U.S. trade deficit and import volume for each partner over the past three years.

Calculate the deficit ratio.

Choose a scaling factor (e.g., 50 %).

Apply the formula.

Publish the new tariff schedule with an annual adjustment mechanism.

U.S. “equivalent‑tariff” algorithm

The Office of the U.S. Trade Representative publishes the formula:

Tariff rate change = Trade deficit ÷ Total imports

The administration then halves the result (multiply by 50 %) as the actual rate.

Applying 2024 official trade data yields numbers almost identical to Trump’s announced tariffs.

Other AI models produce the same formula

Queries to DeepSeek v3, Grok, Gemini all returned the same linear “deficit ÷ imports” core.

Base rate = Trade deficit / U.S. imports of that country; if below 10 % impose 10 %, otherwise use the calculated rate.

Is this mere coincidence?

The analysis compares the AI output with the official USTR proposal, highlighting identical core metric, linear structure, scaling parameter, and minimum rate.

Are AI models becoming “parallel advisors” for policymakers?

The similarity suggests that AI‑driven quantitative reasoning may be influencing modern policy design, even if not explicitly cited.

While we cannot prove that Trump’s team used ChatGPT, the convergence of logic warrants deeper reflection on the role of algorithmic advice in governance.