Frühere Suchanfragen

Keine früheren Suchanfragen

Suchoptionen

Nur verfügbar, wenn angemeldet.
mastodontech.de ist einer von vielen unabhängigen Mastodon-Servern, mit dem du dich im Fediverse beteiligen kannst.
Offen für alle (über 16) und bereitgestellt von Markus'Blog

Verwaltet von:

Serverstatistik:

1,3 Tsd.
aktive Profile

mastodontech.de: ÜberStatusProfilverzeichnisDatenschutzerklärung

Mastodon: ÜberApp herunterladenTastenkombinationenQuellcode anzeigenv4.5.0

#aidetection

1 Beitrag1 Beteiligte*r0 Beiträge heute
Doug Holton

Workado promoted its AI Content Detector as 98% accurate in detecting whether text was written by AI or a human. However, independent testing showed that the accuracy rate for general-purpose content was only 53%. The FTC alleges that Workado violated the FTC Act because the 98% claim was "false, misleading, or non-substantiated." nitter.net/LuizaJarovsky/statu
More on the numerous problems and issues with AI detectors: dangerousaideas.substack.com/p

NitterLuiza Jarovsky, PhD (@LuizaJarovsky)🚨 BREAKING: AI enforcement is on the rise! The Federal Trade Commission issued an order against an AI company over false, misleading, or unsupported advertising. Here's what you need to know: The Federal Trade Commission (FTC) issued an order against Workado requiring the company to stop advertising the accuracy of its AI content detection products UNLESS it has reliable evidence showing the products' accuracy rate claims. According to the FTC, Workado promoted its AI Content Detector as 98% accurate in detecting whether text was written by AI or a human. However, independent testing showed that the accuracy rate for general-purpose content was only 53%. The FTC alleges that Workado violated the FTC Act because the 98% claim was "false, misleading, or non-substantiated." Under this FTC order, the Workado: - Is prohibited from making any representations about the effectiveness of any AI content detection product unless it is not misleading, and the company has competent and reliable evidence to support the claim at the time it is made; - Is required to retain evidence to support such efficacy claims; - Must email eligible consumers about the consent order and settlement with the Commission (see the email template below, attached to the FTC order); - Must submit compliance reports to the FTC one year after the order is issued and every year for the following three years. - My first comment on this case is that there are probably many other AI companies inflating their accuracy rates to seem better than their competitors. The name of this practice is "AI washing," and the FTC is watching. While the EU AI Act is not fully enforceable yet, the FTC and other U.S. agencies are already starting to take action against AI companies. As I've written a few times before, it's not because the U.S. does not have a comprehensive federal AI regulation (like the EU AI Act) that it cannot act against abusive AI practices. It definitely can, and the FTC has been proactive in this area. A last caveat is that, according to the U.S. AI Action Plan, the federal government can: "(...) Review all FTC final orders, consent decrees and injunctions, and, where appropriate, seek to modify or set-aside any that unduly burden AI innovation." So this brings an extra degree of uncertainty to U.S. AI policy and how effective enforcement will be. I'll keep you posted! - 👉 Never miss my curations and analyses on AI: join my newsletter's 73,400+ subscribers (link below).
ResearchBuzz: Firehose

University of Portsmouth: New software will help combat deep fake image threats to personal security. “Realistic images created by artificial intelligence (AI), including those generated from a text description and those used in video, pose a genuine threat to personal security. From identity theft to misuse of a personal image, spotting what’s real and what’s fake is getting harder and […]

https://rbfirehose.com/2025/03/05/university-of-portsmouth-new-software-will-help-combat-deep-fake-image-threats-to-personal-security/

ResearchBuzz: Firehose | Individual posts from ResearchBuzz · University of Portsmouth: New software will help combat deep fake image threats to personal security | ResearchBuzz: Firehose
Mehr von ResearchBuzz: Firehose
#ai#aidetection#aigenerated
*
Carl T. Bergstrom

Numerous sources claim that TurnItIn has a generative AI detector with something like 98% sensitivity and 99% specificity.

For example, the Washington Post, below.

This is completely implausible given that OpenAI themselves only claim to be able to achieve 26% sensitivity and 91% specificity.

So where does this wild claim come from? I think I've figured it out.

washingtonpost.com/technology/

# #

AngesagtLive-Feeds
Über