Post SESTA-FOSTA: AI Enables Freedom of Speech by Understanding Context

There are a *LOT* of articles about how passing SESTA-FOSTA, a bipartisan bill designed to hold websites accountable for their part in human and sex trafficking, is an attack on freedom of speech.

The argument against SESTA-FOSTA, pretty much boils down to sites are going to ban words that could be used in sexual, or slave ads, and opt too far on the side of “safe” such that free speech will be impaired.

10 years ago, even 3 years ago, this would be the case. Today however there are NLP/NLU technologies which can determine if, “When responding please include the Age, Sex, and Breed of the puppy you are wishing to sell” is not about selling sex. More importantly, these systems can also tell that “If your puppy has special skills such as cooking, dancing, or massage, please include that as well,” that it is not an ad for a puppy.

Moderators at companies like Topix, and Reddit, have often been able to keep communities relatively safe, but Craigslist, Backpage, and classified sites have mostly relied on keywords. Hobby and enthusiast sites, like Jeep Forums, and Weight Lifting Forums often have no moderators and simply expect that spammy posts will die.

Recognant has been working for more than a year providing leads from the internet to law enforcement to track down human traffickers. Many of the ads selling a human don’t mention anything human at all. They look like an ad for a used bike, or car. Some don’t look like much of anything, but use a “secret” made up word that makes them possible to be found in Google even if they are spread across 20 random forums, or comment sections.

A site owner could literally see a random comment that says:

I have a 12 year old parakeet for sale that speaks some English. Imported from China, with large plumage on the front, and long legs. Very friendly, and very domestic. Would anyone like to adopt her?

-Ghyzange

And this seemingly innocuous message is actually less coded than many of the messages we find. Decoded this message says that the author has a 12 year old girl from China with large breasts, and is likely quite tall. Can cook, and doesn’t resist sex. To make the ad possible to find the author signs with a name that is not found anywhere else on the internet.

Recognant is able to track down these kinds of ads because we have indexes of enough of the internet to notice when a new word is introduced, and if that word is associated with suspect activity. (Like a comment that doesn’t match the main content)

Recognant’s AI also sees through euphemisms, and had collected more than 10,000 euphemisms used in sexual ads. For example a common differentiator between a legit massage ad, and one for a prostitute can be innuendo, just plain saying something that is not allowed at a legit place.

Our massage therapists are very open-minded. Feel free to enjoy your massage free of the towel if you choose, and enjoy a warm assisted shower after.

None of the words in the above copy say that the massage place is going to touch your penis, nor that they are going to hop in the shower naked with you, but that is what is implied. None of the words in the copy couldn’t be in a perfectly legitimate ad, it is only in context that they show an ad is for sex.

Our massage therapists can do wonders if you are open-minded to new treatments such as toning. Hot Sauna and Warm Shower are available so you can refresh before returning to home or work, or if you are coming from the gym. Sauna, Shower, and Towels are Free. We have assisted countless clients with post workout recovery.

The above ad uses all of the same words, but is completely legitimate, with no innuendo.

Recognant’s AI can determine which is “safe” and which is “unsafe” copy. The definition of “safe” can even be a bit malleable. An ad for a provider could be slightly flirty, and not sexual in some mediums, and totally straight in another depending on the guidelines of the site.

All of this is possible because Recognant uses advanced psychographic analysis provided by an epistemology. A very specific type of AI which was invented at Recognant. That isn’t to say Recognant is the only company that could offer such a service, but they are the only ones doing so affordably. IBM Watson contains most of the necessary parts to do similar things, but pricing would exceed what a site moderator costs. SRI could similarly have an offering, but again it would be cost prohibitive.

The challenge with SESTA-FOSTA really is that many small sites may have to choose between allowing users to comment and complying with the law. Previously there was a “Safe harbor” provision that said that if a site reasonably was unaware that their site was being used they were safe, and that simply taking down flagged content was good enough. Moving the onus to the site owner, and making it proactive rather than responsive, means every person with a blog could be at risk (and they are).

Recognant working to make its solution affordable to as many as possible. Our initial offering is $300 a month. This is $10 a day, which is about what one would pay to have someone spend an an hour a day checking that all the comments were safe. This is far cheaper than a moderator, and unlike a moderator who is likely to only see posts from a single site, our AI is using data from 1000s of sites to spot the new trends and tricks used by those looking to avoid detection.

To find out more about Recognant’s AI based SESTA-FOSTA content moderation visit www.sesta-fosta.com

#EndTrafficking #Sesta #Fosta #AI #Legal #Censorship