Mozilla Adapts 'Fakespot' Into an AI-Detecting Firefox Add-on (omgubuntu.co.uk) 36
An anonymous reader shared this post from the blog OMG Ubuntu
Want to find out if the text you're reading online was written by an real human or spat out by a large language model trying to sound like one? Mozilla's Fakespot Deepfake Detector Firefox add-on may help give you an indication. Similar to online AI detector tools, the add-on can analyse text (of 32 words or more) to identify patterns, traits, and tells common in AI generated or manipulated text.
It uses Mozilla's proprietary ApolloDFT engine and a set of open-source detection models. But unlike some tools, Mozilla's Fakespot Deepfake Detector browser extension is free to use, does not require a signup, nor an app download. "After installing the extension, it is simple to highlight any text online and request an instant analysis. Our Detector will tell you right away if the words are likely to be written by a human or if they show AI patterns," Mozilla says.
Fakespot, acquired by Mozilla in 2023, is best known for its fake product review detection tool which grades user-submitted reviews left on online shopping sites. Mozilla is now expanding the use of Fakespot's AI tech to cover other kinds of online content. At present, Mozilla's Fakespot Deepfake Detector only works with highlighted text on websites but the company says it image and video analysis is planned for the future.
The Fakespot web site will also analyze the reviews on any product-listing pages if you paste in its URL.
It uses Mozilla's proprietary ApolloDFT engine and a set of open-source detection models. But unlike some tools, Mozilla's Fakespot Deepfake Detector browser extension is free to use, does not require a signup, nor an app download. "After installing the extension, it is simple to highlight any text online and request an instant analysis. Our Detector will tell you right away if the words are likely to be written by a human or if they show AI patterns," Mozilla says.
Fakespot, acquired by Mozilla in 2023, is best known for its fake product review detection tool which grades user-submitted reviews left on online shopping sites. Mozilla is now expanding the use of Fakespot's AI tech to cover other kinds of online content. At present, Mozilla's Fakespot Deepfake Detector only works with highlighted text on websites but the company says it image and video analysis is planned for the future.
The Fakespot web site will also analyze the reviews on any product-listing pages if you paste in its URL.
Bullshit (Score:3, Interesting)
This is proof Mozilla needs to be disbanded and Firefox put under proper management.
If the Mozilla foundation had found a way to tell what's "AI generated text", this would be generating headlines world wide and they'd be promoting it as an incredible breakthrough, not some bullet-list feature on a fucking plug-in. Literally nobody else has been able to do this, and students across the world (for example) are currently getting dicked over by "AI detection tools" created and promoted by scammers that do not work and never have done.
Human beings, alas, will continue to be the only people who can determine whether text is *suspicious* or not, but even then that's the only judgement we can make. Oh, it uses em-dashes? The story contains the phrase "Since them my family has been blowing up my phone"? Well, there's a good chance it's AI generated. But that's all we can do at this stage.
The current management of Mozilla needs to resign. They've turned the most important open source project into a harbor for scammers. They *are* scammers. We need to take Firefox away from them.
Re: (Score:3)
The word you are looking for is "fork".
Some organization who know how to maintain large open source projects would need to fork Firefox and maintain it in a way that it can survive even when Mozilla would abandon it. I see GNU and the FSF in the position to organize something like that. GNU already has a Firefox fork, even though it is limited by defaulting to an free-software-only addon repository.
Re: (Score:2)
Human beings, alas, will continue to be the only people who can determine whether text is *suspicious* or not, but even then that's the only judgement we can make.
Judging by recent political events, I'm dubious humans can do even that, especially when it conflicts with what they want/hope to be true.
Re: (Score:2)
Judging by recent political events, I'm dubious humans can do even that
Not all humans, but some humans can tell.
But, the way that Generative-AI works, if an AI can tell that text was written by an AI, then it can learn to write text that can't be distinguished by an AI.
Re: (Score:1)
Human beings, alas, will continue to be the only people who can determine whether text is *suspicious* or not, but even then that's the only judgement we can make.
Judging by recent political events, I'm dubious humans can do even that, especially when it conflicts with what they want/hope to be true.
True; imagining that your preferred politics are a reality-detector is pretty good evidence of poor judgment.
Re: (Score:2)
Human beings, alas, will continue to be the only people who can determine whether text is *suspicious* or not, but even then that's the only judgement we can make.
Judging by recent political events, I'm dubious humans can do even that, especially when it conflicts with what they want/hope to be true.
True; imagining that your preferred politics are a reality-detector is pretty good evidence of poor judgment.
More to the point, ignoring obvious lies and dis/misinformation is good evidence of poor judgement. One political party recently seems to rely on people doing that more than the other. Why their supporters do that may have many facets, but it's still poor judgement.
Re: (Score:2)
Re:Bullshit (Score:4, Insightful)
Re: (Score:3)
If the Mozilla foundation had found a way to tell what's "AI generated text", this would be generating headlines world wide and they'd be promoting it as an incredible breakthrough, not some bullet-list feature on a fucking plug-in.
What the fuck are you talking about? The methods of detecting AI generated text are quite well known and methods have been published into research papers for over a year now. Mozilla hasn't created anything here. They've simply taken an open source detection model https://github.com/LLNL [github.com] and put it into a browser plugin.
Chill out man, you'll work your way up to a stroke at this rate.
Re: (Score:2)
This is the kind of stuff that Mozilla should be doing. Just like they have warnings for fake phishing sites, they should have warnings for fake AI reviews and fake AI generated images. If they can develop technology to spot that stuff it will be valuable to users.
Fakespot is a good add-on, useful for quickly checking things like Amazon reviews to see if the product was heavily promoted by the seller and thus likely to be of very poor quality.
Re: (Score:2)
They should also be working on clean coal, table top fusion, and a washing machine that doesn't lose socks too.
The fact the application would be useful if it had a chance in hell of working doesn't mean that it works. And all it's doing ultimately is contributing to the lie that AI can detect AI generated content, which leads to people being kicked out of college for cheating on the basis of the same logic being applied to the content they themselves worked on.
Do you seriously not see the fucking problem wi
Re: Bullshit (Score:1)
Re: Bullshit (Score:2)
Too bad you didn't add them manually using the HTML entity, then slashdot would have rendered them.
Re: Bullshit (Score:2)
You're not wrong.
There are many ways in which slashdot falls short of standards, especially under current leadership.
What Reviews? (Score:5, Insightful)
Reviews are being replaced by AI review summaries just like how Question sections have disappeared. You can still get to those on Home Depot, but Amazon has completely neutered that feature of their store rendering it near useless.
Further, Amazon is currently promoting video reviews but those are actually paid ads. If you're a high enough influencer you can be paid to submit video reviews and you get a commission if someone watches your review prior to buying the product. Have you noticed how the amount of video reviews for products have exploded and they all just hold up the product and just say "I liked it"? That's why. The recommendation for if you get into this program is to do quick video reviews for everything in your house that Amazon sells. Then visit a friend and take inventory of their house. Then when you run out of friends, you order random stuff and turn them after you've finished recording. At least the video ads give you more product angles than the product images. Seriously, it's 2025 and a ton of companies still aren't putting pictures of every side of their product on the product pages. WTF businesses? Why do you make it harder for us to buy your products? Having full product coverage was a think known back when eCommerce first got started. The customer can't pickup the product when shopping online so we've go to be sure to include all of it in images. I don't know why businesses stopped caring about doing that.
I could go on ranting about reviews and products not matching, selling fraudulently advertised products (lookup vids for false fuses specs) and a bunch of other things, but instead I'll take a deep breath and shut up now. At least I've been able to cut the amount of stuff I buy from Amazon in half. Do I want to support that anti-user company or do I want to spend an extra 2 hours looking for items and receiving 12 boxes in the mail instead of one?
Re: (Score:2)
>"Reviews are being replaced by AI review summaries just like how Question sections have disappeared. You can still get to those on Home Depot, but Amazon has completely neutered that feature of their store rendering it near useless."
Thank you. I am glad I am not the only one very annoyed by this.
Yes, the AI thing has been helpful sometimes. But most times I want to search and decide what I want to read from the reviews and Q/A. So now you have to scroll all the way down the reviews, then click on "mo
Seems to me it runs on their cloud (Score:4, Informative)
Though the 4 engines called by the extension have code on github, the extension stops working if I pull the Ethernet cable (or stop network through the system scripts, in my case /etc/init.d/net.enp4s0 stop ) . So I guess the engines are open source, but the training data are at mozilla.org. Which would be fine but they don't make clear that the text under analysis is being sent outside.
Re:Seems to me it runs on their cloud (Score:5, Informative)
They make it clear when you install the add-on. It explains your highlighted text will be collected and sent and it requires you to opt-in before it can be used.
Re:Seems to me it runs on their cloud (Score:4, Informative)
Thanks, my bad. I got distracted by the "help us improve" checkbox, I thought the opt-in was about that. Though I think it should have been made clear before decision point of downloading.
Re: (Score:2)
The training data isn't hosted at mozilla.org. It's hosted on Github and run locally. Only results are sent for analysis. This isn't running in the cloud, and there's no reason for it to be.
Re: (Score:2)
The addon itself says it will send the text to be analyzed. Upon installation, it says:
Deep Fake Detector will send highlighted text to be analyzed when you request analysis by any method.
If you choose to not allow us to send this data, we will not be able to analyze text.
Then there are only two options "I opt in" and "I opt out". If you press "I opt out", it uninstalls the add-on.
unfortunately (Score:2)
many AI models were trained on my writing style, so my work will be flagged as AI :(
Fix *real* things, don't hop on the AI fadwagon (Score:2)
Firefox has become a memory hog and still has a lousy spell-checker: century-old Soundex would often be better (have it use both).
Better use case(s) (Score:2)
It would be much more helpful and time-saving if something like this (provided it works as advertised) could be incorporated into web search engines and don't even show bot-generated pages. Knowing that search engines and bot pages both get revenue from advertising, I doubt we'll ever see it though. Perhaps one day I or someone else will come up with a Tamper Monkey script that tampers with the results page...
Another tangent that comes up is a script that uses a chatbot to summarize the heck out of these l
Local or not? (Score:2)
Re: (Score:1)
That second one. Like when they acquired Pocket or Anonym, this is the Mozilla foundation finding new ways to monetize Firefox by having it collect your browsing information.
meh (Score:2)
Plot twist: it works only in English