The AI hype is real, and some companies are facing consequences for misleading claims.
Today, the Federal Trade Commission (FTC) ordered a company called Workado to "back up" claims that its tool to detect AI-generated text is "98% accurate." According to independent testing, Workado's tool is only 53% accurate, the FTC says. It was also only trained to "classify academic content," despite being advertised to work on any online content.
"Consumers trusted Workado's AI Content Detector to help them decipher whether AI was behind a piece of writing, but the product did no better than a coin toss," says Chris Mufarrige, Director of the FTC's Bureau of Consumer Protection. "Misleading claims about AI undermine competition by making it harder for legitimate providers of AI-related products to reach consumers."
Workado operates under the name BrandWell. It claims its "AI Detector" was one of the first on the market in 2022 and works with text generated by ChatGPT, Gemini, Claude, and other models.
"Most free AI detectors and classifiers are outdated and scan for old language generators," an FAQ says. "'AI Detector' identifies all versions of GPT, including GPT-4, Claude Sonnet, and Google Gemini. It does this by comparing human vs LLM written articles, essays, and more combined with understanding of LLM models and how words and sentences are constructed to identify probably word choices and sentence structures."
However, disclosures at the bottom of the website indicate the company can't guarantee accuracy. BrandWell notes that its detector can have "varying levels of accuracy" and recommends people use discretion when reviewing the findings. Identifying prose from a computer versus human is difficult, even for OpenAI, which shut down its AI text-detection tool in 2023 due to inaccuracies.
Incorrect assessments can have devastating effects, particularly on students. Though AI-powered cheating is a concerning and growing trend, teachers and administrators need accurate tools to combat it effectively. BrandWell's website acknowledges the risks by urging users to "openly communicate with those you're working with, if you suspect the use of language models, then engage in a dialogue to address the issue ethically."
From now on, the FTC is banning Workado from making claims about its tool's effectiveness unless it has "reliable evidence." The company must also email eligible consumers about this news, and submit compliance reports to the FTC every year for three years.
Earlier this month, federal prosecutors in New York charged 35-year-old Albert Saniger for claiming his company used AI when it was actually humans pushing buttons behind the scenes. He raised $40 million for the e-commerce company, called Nate, and now faces prison time.
Apple has also been sued three times for false advertising related to Apple Intelligence features and its availability on the iPhone 16. The Better Business Bureau's National Advertising Division weighed in last week to recommend that Apple be more transparent about what's available to consumers now.
