accessiBe; Analysis of Proposed Consent Order To Aid Public Comment

What this rule actually says

The FTC charged accessiBe with making false claims about how well its AI accessibility tools work and failing to deliver the promised features. This consent order requires accessiBe to stop making unsubstantiated claims about product capabilities and to actually deliver what's advertised. The core issue: saying an AI tool does something it doesn't actually do is an unfair or deceptive practice.

Who it applies to

• If you market AI capabilities you haven't actually built or tested: This applies to you, regardless of jurisdiction or product type.
• If you claim accessibility features (WCAG compliance, disability support, etc.) without evidence: Medical scribes, hiring assistants, and support chatbots making these claims are directly in scope.
• If you're in the US: The FTC's authority covers US-based companies and those serving US customers.
• If you're collecting or using any user data to train or improve your model: The order emphasizes substantiating claims about how data is used; making vague privacy promises without backing them up is risky.
• If you target healthcare, employment, or accessibility-sensitive use cases: These verticals face extra scrutiny for unsubstantiated claims.

What founders need to do

1. Audit your marketing claims (1–2 days): Go through your website, pitch deck, demo videos, and customer-facing materials. Flag any statement about what your AI does or how well it performs. Is each claim actually true based on your testing?

1. Document your evidence (3–5 days): For any claim about accuracy, speed, or capability, gather the testing data that backs it up. "We tested with 100 customer calls" beats "our system is highly accurate" with nothing behind it.

1. Rewrite marketing conservatively (2–3 days): Replace unproven claims with what you can actually demonstrate. Instead of "eliminates scribing errors," try "reduces manual transcription time by 40% (based on internal testing with 50 calls)."

1. Make privacy and data practices explicit (1 day): If you're training on user data or using data in any way beyond running the product, say so clearly—don't bury it or make vague statements.

1. Monitor and update ongoing: Whenever you add a feature or make a capability claim, make sure you're testing it and can back it up.

Bottom line

If you're making claims about what your AI does without evidence to back them up, the FTC can now point to this order and sue—so audit your marketing language now and only claim what you've actually tested.