VC
Value Add VC
⚡HomePulse⚡Helpful Apps📝Blog
← Value Add PulseREGULATIONComment period through Jul 31, 2026

Why the FTC's AI 'Accuracy' Rule Could Reshape Chatbots

The FTC's proposed policy statement treats undisclosed steering of AI outputs as potential deception under Section 5, singling out Colorado's AI Act by name and opening a comment period through July 31.

Jul 31, 2026
Comment Period Ends
>90%
User Non-Verification Rate
TC
Trace Cohen
Early-stage VC & angel · Founder, New York Venture Partners
July 5, 2026
1 min read
ShareXLinkedInEmail
THE RUNDOWN
1

The FTC's proposed statement argues that secretly steering an AI model's output away from the most accurate answer can constitute deception under Section 5 of the FTC Act

2

It explicitly names Colorado's Artificial Intelligence Act as a law that could pressure companies into suppressing accurate outputs to avoid disparate-impact liability

3

The statement notes that consumers accept AI outputs without independent fact-checking more than 90% of the time, forming the basis for a 'reasonable expectation' of undisclosed-manipulation-free accuracy

4

Companies can avoid liability under the proposed framework with clear, conspicuous disclosures that a system prioritizes objectives other than raw accuracy

TC
The VC Read · Trace's TakeTrace Cohen

Buried in a policy statement about 'accuracy' is a much bigger fight: whether federal deception law or state fairness law wins when they conflict inside a single model's outputs. Any founder building a chatbot, recommender, or moderation tool should start drafting disclosure language now -- 'we didn't tell you what we optimized for' is about to become a much more expensive mistake.

The Federal Trade Commission opened a public comment period on July 1 for a proposed policy statement addressing what it calls the 'suppression of accuracy' in AI systems -- a framework that could reshape how every consumer-facing AI product discloses when it's optimizing for something other than the most accurate answer. The comment period runs through July 31, 2026.

The core legal theory: AI companies have made explicit and implicit representations that their systems are designed to produce the most accurate output possible, and because consumers accept AI outputs without independent fact-checking more than 90% of the time, according to the FTC's own citation, they have a reasonable expectation that outputs aren't secretly being steered toward an undisclosed objective. Steering a model's output away from a correct answer without telling users, the statement argues, can amount to deception under Section 5 of the FTC Act.

“Steering a model's output away from a correct answer without telling users, the statement argues, can amount to deception under Section 5 of the FTC Act.”

A significant portion of the statement takes direct aim at state AI laws, singling out Colorado's Artificial Intelligence Act by name as a law that could pressure companies into altering model outputs to avoid disparate-impact liability -- effectively arguing that state fairness mandates could conflict with a federal accuracy mandate. The proposal does include a safe harbor: a company can avoid Section 5 liability by making clear, conspicuous and adequate disclosures that its system is designed to prioritize certain objectives over pure accuracy.

For any startup building consumer-facing AI products -- chatbots, recommendation engines, content moderation tools -- this creates a new compliance surface: disclosure language about what a model optimizes for becomes a legal requirement, not just a trust-and-safety nicety, and conflicts between state fairness laws and this federal framework could force difficult product design tradeoffs.

What to watch: the volume and substance of comments filed before the July 31 deadline, particularly from state regulators pushing back on the Colorado callout, and whether the FTC moves from a policy statement to actual enforcement action against a specific company before year-end.

ShareXLinkedInEmail

Originally reported by Federal Trade Commission. Analysis and editorial commentary by Value Add Pulse.

← Back to Pulse

THE WIRE in your inbox

Tech, startup & VC news with Trace's take. Free, no spam.

Read Next

REGULATION~$42.6B stake offered

OpenAI Offers US Government a 5% Equity Stake

OpenAI has proposed giving the Trump administration a 5% equity stake worth roughly $42.6 billion, an unusual move Sam Altman is framing as sharing AI's upside but which lands amid mounting political scrutiny of the company.

REGULATIONManslaughter charges filed

Tesla Driver Faces Manslaughter Charges After FSD Crash Killed a Woman Inside Her Texas Home

Michael Butler, 44, was arrested and charged with manslaughter after his Tesla Model 3, which he claimed was using Full Self-Driving, crashed through a home in Katy, Texas on June 19, killing 76-year-old Martha Avila inside. An arrest affidavit shows Butler's phone had multiple Google searches from May 2026 about FSD being 'not aggressive enough,' while Tesla AI head Ashok Elluswamy said on X the driver 'manually overrode self-driving by pressing the accelerator all the way to 100%.'

REGULATION$4.7B fine confirmed

Google Loses Final EU Appeal, Must Pay $4.7 Billion Android Antitrust Fine

The Court of Justice of the European Union dismissed Google and Alphabet's final appeal on July 2, confirming a record antitrust fine over how Google bundled Search and Chrome as defaults on Android. The penalty was originally set at €4.34 billion ($4.9B) in 2018, trimmed to €4.1 billion ($4.7B) by a lower court in 2022, and with no further avenues for appeal, Google now has to pay the full amount.

@Trace_Cohen·t@nyvp.com