Artificial Intelligence Flags Legitimate Email as Spam?!?!

Fed up with losing email to your spam folder? Gmail’s ...

Generally speaking, we can say tech companies go to great lengths to never have to admit when they’ve made mistakes. Such is the case lately with the 2 most popular cloud-based email providers, Google’s Gmail/GSuite and Microsoft’s Office 365. They’re not being transparent about their latest foible, specifically that their latest “improvements” in artificial intelligence are flagging legitimate messages as spam, frustrating many and infuriating most.

While no one has officially reported it, yet, how do we know this is happening? For one, Google is promoting their successes in blocking additional spam using the tech all over the Internet but not reporting their failures. For two, social media has been lighting up with mass reports, as people are wondering why people who they’ve emailed back-and-forth with for years are suddenly ending up in spam folders.

This pattern follows historical issues that either did or didn’t ever come to light in the mainstream media.

Each company has their own flavor of this tech. Google calls it TensorFlow and Microsoft when with the less sexy Cognitive Toolkit or CNTK.  As one might imagine, each has its own pros and cons, including strengths and weakness.

The GOOG and MSFT are aggressively working on and updating these, too, which creates issues on the consumer side, such as the recent rash of massive amounts of false positives or flagging legitimate messages as spam.

This isn’t anything new as anyone who has used these services for years can attest. Anyone curious enough to do some quick research can find this has always been a part of using these services because they aren’t obliged to tell anyone when they make major or even minor adjustments that impact performance. They’ve been doing this long before their latest AI technology ever got involved.

Here are just a few examples from over the years, in no particular order:

https://productforums.google.com/forum/#!topic/gmail/beJsHExyG2E
https://blog.returnpath.com/gmail-flags-nearly-5-percent-of-legitimate-email-as-suspicious/

It’s going to be a bumpy ride for a while as AI comes into the mainstream, as machines who’ve learned how to do everything from who we email regularly, who we don’t, and even how we communicate by emulating our writing style, continuing to learn the way anyone does: by making mistakes. 

This is just one of the many reasons our clients love our cyber resilience approach that offers more advanced email configurations that can help keep these hiccups and more impactful ones to a minimum. Machines aren’t the only ones learning to do things better.