Artificial intelligence is becoming more influential in decisions that can shape a person's financial life—from whether someone qualifies for credit to how job applicants are screened or housing applications are evaluated.

Now, a proposed Federal Trade Commission policy statement has opened a new debate over how companies should handle bias and accuracy in AI systems.

The Federal Trade Commission proposed the policy statement on July 1, 2026. The agency says it is intended to clarify when companies marketing artificial intelligence systems could violate federal law by representing systems as more accurate or objective than they actually are.

The National Consumer Law Center (NCLC) is urging the FTC to withdraw the proposal, arguing that portions of the policy could discourage companies from testing for and correcting discriminatory outcomes. NCLC says that could be especially consequential when automated systems are used in credit, housing, employment and other decisions that materially affect consumers.

The disagreement highlights a larger challenge facing regulators: how should companies measure AI "accuracy" when the data used to train or operate those systems may already reflect historical inequalities?

What the FTC Is Proposing

The FTC's proposed Policy Statement Concerning the Suppression of Accuracy in Artificial Intelligence Systems focuses on Section 5 of the Federal Trade Commission Act, which prohibits unfair or deceptive acts or practices.

The Commission's concern is primarily about how AI products are marketed.

If a company promotes an artificial intelligence system as accurate, neutral or reliable, the FTC argues that intentionally altering its outputs for reasons unrelated to predictive accuracy could make those marketing claims misleading in some circumstances.

The proposal does not create a new federal AI law. Instead, it explains how the FTC believes its existing consumer-protection authority could apply to representations made by companies selling or deploying AI systems.

That distinction matters because the debate is not simply over whether AI should be "biased" or "unbiased." It centers on how accuracy is defined, how bias mitigation affects model performance, and what companies should disclose to consumers about those tradeoffs.

Why Consumer Advocates Are Pushing Back

NCLC argues that the FTC's framework oversimplifies the relationship between bias and accuracy.

According to the organization's August 4 response, efforts to identify and correct discriminatory patterns can actually make AI systems more accurate for populations that were poorly represented or historically disadvantaged in the underlying data.

The concern is that an AI model can appear statistically accurate overall while performing significantly worse for particular groups.

If developers are discouraged from testing and correcting those differences, consumer advocates argue that inaccurate or discriminatory outcomes could become embedded in automated decisions.

This is especially important because historical data is not automatically neutral. Past lending, employment, housing and insurance outcomes can reflect earlier discrimination or unequal access, meaning AI trained on those records may reproduce some of the same patterns.

NCLC has previously warned that financial institutions using machine learning can gain efficiency while also creating risks involving unlawful discrimination and unfair or deceptive practices.

Credit Decisions Show Why the Debate Matters

Credit underwriting is one of the clearest examples of how AI can directly affect household finances.

Traditional lending decisions may consider factors such as income, debt, repayment history and credit scores. More sophisticated automated models can analyze significantly larger datasets and identify patterns that human underwriters might miss.

That can potentially make lending faster and expand access to some borrowers.

But it can also introduce new risks.

An algorithm might rely on data points that appear neutral but strongly correlate with race, geography, income or other characteristics. Models trained on historical lending decisions may also learn patterns shaped by past discrimination.

In practice, that could influence:

• Whether someone is approved for a loan

• The interest rate they receive

• Their credit limit

• Whether additional documentation is required

• How lenders assess repayment risk

An automated decision does not become fair simply because a computer made it.

The central regulatory question is whether companies are adequately testing these systems for both accuracy and discriminatory outcomes.

Housing Decisions Are Increasingly Automated

Artificial intelligence and automated screening systems are also becoming part of housing decisions.

Landlords and property managers may use technology to evaluate rental applications, verify income, assess risk or screen prospective tenants.

Mortgage lenders can also use automated underwriting tools when evaluating borrowers.

These systems can make processing faster, but errors can carry substantial consequences. Incorrect information, poorly designed models or biased data can potentially affect whether someone obtains housing or what financing terms they are offered.

That makes transparency particularly important.

Consumers may not always know which automated tools influenced a decision, what information those tools considered or how to challenge an inaccurate result.

The FTC's broader history with artificial intelligence already reflects concerns about inaccurate automated systems. In 2023, for example, the agency brought an enforcement action against Rite Aid over its use of facial recognition technology, alleging that inadequate safeguards produced false-positive matches that harmed consumers.

Employment Adds Another Layer of Financial Risk

The implications extend beyond borrowing and housing.

Employers increasingly use automated systems to help screen résumés, assess candidates, rank applicants or evaluate employees.

A biased or inaccurate system could therefore affect someone's ability to earn income in the first place.

That creates a direct connection between AI regulation and financial wellness.

A consumer denied a mortgage because of a flawed model suffers one type of financial harm. A worker screened out of employment opportunities by an inaccurate algorithm may face another.

Both situations illustrate why automated decision-making is increasingly becoming a consumer-finance issue rather than merely a technology issue.

Federal Agencies Have Previously Warned About Automated Discrimination

The current FTC proposal also represents a notable evolution in how federal regulators are discussing artificial intelligence.

In 2023, the FTC joined the Consumer Financial Protection Bureau, Department of Justice and Equal Employment Opportunity Commission in emphasizing that existing civil-rights and consumer-protection laws continue to apply when companies use automated systems.

At the time, federal officials warned that AI could amplify fraud and automate discrimination and emphasized that companies do not receive an exemption from existing law simply because a decision is made using new technology.

The FTC has separately acknowledged risks involving AI bias, discrimination, privacy and deceptive practices in previous research and enforcement work.

The 2026 proposal does not erase those laws. But consumer groups worry that its framing of efforts to modify AI outputs could create uncertainty about how businesses should simultaneously satisfy accuracy expectations and anti-discrimination obligations.

Accuracy and Fairness Are Not Always Opposites

One of the most important issues in this debate is the assumption that improving fairness necessarily means sacrificing accuracy.

That relationship can be much more complicated.

If an AI model is built using incomplete, unrepresentative or historically distorted data, correcting those weaknesses may improve performance rather than degrade it.

For example, a credit model that performs well for borrowers represented heavily in its training data but poorly for another population cannot necessarily be described as equally accurate for everyone.

Evaluating AI therefore requires more than asking whether the system produces the correct prediction most of the time.

Developers may also need to examine:

• Which populations were represented in the training data

• Whether error rates differ significantly among groups

• Which variables influence decisions

• Whether proxy variables reproduce prohibited discrimination

• How consumers can challenge inaccurate outputs

• Whether humans review consequential automated decisions

Those questions become particularly important when an algorithm can determine access to credit, housing or employment.

Consumers May Not Know AI Is Involved

Another challenge is visibility.

When someone applies for a loan, apartment or job, they may interact with a familiar company while an outside technology provider performs much of the underlying analysis.

That can make it difficult to understand why a particular decision occurred.

Consumers may receive a denial without knowing whether an algorithm influenced it, which data was used or whether that information was accurate.

Existing laws can require explanations in some settings—for example, lenders generally must provide specific reasons for adverse credit decisions—but increasingly complicated algorithms create new questions about how meaningful those explanations are.

AI regulation will therefore need to address not only how models are built but also how their decisions are communicated.

What Consumers Can Do Today

The FTC's proposal remains a policy discussion rather than a final rule governing every automated decision.

Consumers do not need to become AI experts, but they can take practical steps when a financial or employment decision appears incorrect.

If you receive an unexpected denial or unfavorable decision:

• Review the explanation provided.

• Verify that your credit reports contain accurate information when credit is involved.

• Request correction of inaccurate personal data.

• Keep copies of applications, notices and correspondence.

• Ask how to dispute or appeal the decision.

• Document any information you believe was incorrectly considered.

• Report suspected discrimination or deceptive practices to the appropriate regulator.

The growing role of artificial intelligence makes maintaining accurate financial records and personal information increasingly important.

AI Can Be Powerful Without Replacing Human Judgment

The debate over the FTC proposal reflects a larger question that will continue as AI becomes more embedded in financial services.

Artificial intelligence can process enormous amounts of information, identify patterns and help professionals make decisions more efficiently.

But an algorithm is only as reliable as its data, design, testing and oversight.

When decisions affect someone's ability to obtain a home, borrow money or earn income, efficiency cannot be the only measure of success.

Consumers also need accuracy, transparency and meaningful protections when automated systems get something wrong.

The challenge for regulators and businesses will be creating standards that encourage innovation without allowing technology to obscure accountability.

Keep Your Financial Information Clear and Organized

As more financial decisions rely on data, the accuracy of your own financial information becomes increasingly important.

Copiafy helps you organize important financial documents, monitor your credit, track goals, manage bills and keep a clearer view of the information shaping your financial life.

Technology should help you understand your finances—not make them harder to navigate.

Create your free Copiafy account and build a more organized financial picture in one secure workspace.