by Debbie Burke
Recently I had a disturbing email experience.
For some months, circumstances had prevented the five members of my critique group from meeting face to face. So we began exchanging group emails to bring each other up to date.
Since we’re friends as well as writing colleagues, our emails often include personal information about families, friends, dogs, health, etc.
With five people chiming in, a recent email chain became quite long.
Then one member received a pop-up notice at the top of her gmail that gave an “AI Overview” summarizing each person’s contributions to the discussion.
Where the &$%# did that come from??? How did a bot gain access to our emails?
Our conversations included deeply personal medical information about ourselves, family, and friends such as…
Who’s struggling with symptoms that doctors can’t diagnose? Who needs heart or brain surgery? And so on.
Private, personal, confidential conversations among close friends.
Out of nowhere, an AI bot gave us a nice, neat, efficient, accurate summary.
How helpful. But intrusive as hell.
How did this nosy bot access, read, and summarize our discussions?
Had an update from Gmail changed settings to allow AI summaries?
Click the following link for an article from HuffPo that describes what probably happened and reasons why we might not want a nosy little bot to read our emails.
More insights from Proton.me:
“Today, companies like Google are expanding AI access to private communications such as email, framing it as productivity and convenience. But Gemini operates under its own terms, making it harder to distinguish what data is handled by Gmail itself and what is processed by AI systems.”
If you don’t want Gemini AI summaries on Gmail, here’s how to change “smart” settings: help page.
When I checked my settings, I had already turned off “smart” features. Yet the AI summary still showed up. Hmmm.
That leads me to believe someone else hadn’t disabled their smart features, which opened access to our Gmails.
***TKZ’s tech experts, please feel free share your knowledge in the comments.***
What does that mean for medical and legal professionals who send and receive confidential records? If a recipient doesn’t know to shut off their device’s smart features, can Gemini suck up private information for its own commercial use?
Doesn’t that violate HIPAA rules and attorney-client confidentiality???
I foresee class action lawsuits from victims damaged by confidentiality breaches.
What about writers?
We routinely email manuscripts to agents and editors. We also exchange manuscripts for beta reading, critique, editing, etc. Those manuscripts are copyrighted as soon as the author commits them to tangible form, on paper, digital file, etc. That protects our work, right?
Not necessarily.
You may have heard about the $1.5 billion judgment against Anthropic for using illegally obtained copyrighted books to train Claude, their large language model (LLM) AI program.
The award was a win for authors, right? Uh, only under limited conditions.
To qualify for compensation in the Anthropic settlement, their books had to be registered with the US Copyright Office, not just copyrighted.
Typically, traditional publishers register copyrights but some companies didn’t. Their authors were out of luck.
Also typically, copyrights are registered upon publication, after edits, rewrites, additions, etc.
That leaves many manuscripts in limbo.
What if we email manuscripts to agents or editors? Our work is copyrighted but, while it’s under submission, it’s probably not yet registered. Can these be vacuumed up to train LLMs?
Currently, regulation of AI’s use is virtually nonexistent. Laws haven’t caught up with constantly changing developments. Legislation to control and limit use is likely years away, maybe even decades.
Meanwhile, the ease, convenience, and efficiency of technology has seduced us into giving up privacy and confidentiality.
I turned off annoying Gemini intrusions by changing settings on my own computer, but I can’t control others’ devices. And of course I trust Google as much as that nice Nigerian prince who’s sending me millions.
Yes, I could switch to a different email server but that would cut off my main contact point as an author.
I don’t know how to deal with this except to be more cautious of what I write in emails.
Back in 2019, I wrote about text messages that I naively thought were private. Then I learned Facebook, Amazon, Google, etc. had accessed my texts to send advertising related to them. Stealth permissions buried deep in the phone’s terms and conditions grant access to third parties. By using the phone, you agree to the conditions, even when they’re next to impossible to find.
Six years later, Gmail is in a similar state where the onus is on the user to go extra miles to opt out of invasions into privacy.
This reminds me of wise advice from an attorney mentioned in the 2019 post: “Don’t put in writing anything you wouldn’t want to be read in open court.”
~~~
TKZers: Have you run into Gemini’s email summaries? What do you do to maintain online privacy? Or does that no longer matter?
~~~
You can’t believe your eyes. Can investigator Tawny Lindholm and attorney Tillman Rosenbaum save an innocent woman’s life after deep fake videos show the world she’s guilty? Find out in Deep Fake Double Down, winner of BookLife’s best mystery contest.
When the AI summary appeared in my gmail, I searched for how to turn it off. At the time I was annoyed that the summary wasn’t any shorter than the original message. It also made me feel stupid like I didn’t know how to read my own messages. I hadn’t thought about other people not turning their “smart” features off and that would lead to my messages still being read. Sigh, the machine bots are taking over the world.
Alec, agreed.
Making you “feel stupid” (which you definitely are not!) could be part of the strategy to convince humans to depend on the machines.
Pingback: Emerging privacy concerns as AI summaries infiltrate Gmail conversations | Noah News
When I got the PDF of my latest novel from my publisher for a final proofread, Adobe sent me a little AI message: This looks like a very long document. Would you like me to summarize it for you?
Sharon, thank you for providing an actual example. This verifies my suspicion that not-yet-published manuscripts are at risk of AI scraping.
Best wishes on your upcoming book launch!
I got the same message from Adobe.
This is a very real issue. I noted that your first source was proton.me. Proton offers free secure email. By secure, I mean supposedly secure enough for whistle blowers in places like China, and Iran. Their servers are in Switzerland. If you want to hide your communications from law enforcement, proton is a good place to start.
Yes, AI is reading your email. It may not be attaching your name to your data, but it does know that you asked Aunt Martha how to get the morning after pill.
For creatives, it has read your proposal and rough draft, and first edit. Some of it is for system training. Some for data mining. But chances are I could ask an AI system to write a story in your voice with your characters and get a passable result.
Wow, Alan, thanks for that additional info on Proton.me. Amazing there is still a secure way to communicate.
BTW, after I asked Aunt Martha’s advice, I started receiving ads for the morning-after pill 😉
Scarier is AI for “research”. I know a religious leader who is proud of his ChatGPT generated sermons. About every third one includes a Biblical quote that AI created as needed.
Every week several attorneys are called before the bench because they have filed AI court papers with fictitious sites. There are a few companies that review every court document looking for these instances. They find a few a day.
Therin is AI’s greatest flaw. It will create an answer. If it needs to create sources to back up its answer, OK.
Alan, even scarier is medical research. Recd this morning: https://open.substack.com/pub/garymarcus/p/please-dont-trust-your-chatbot-for?utm_campaign=post&utm_medium=email
As Gary says, “frequently wrong, never in doubt.”
I use a Thunderbird account for my professional email and haven’t noticed any offers to summarize.
I have a gmail account for personal communications and might have already turned off AI ‘helpers’ because I’m not seeing much there, either.
Yet.
“Yet” is the operative word, Terry.
We’ve been binging on Person of Interest. filmed from 2011-16. Prescient.
One more. If your WIP includes a cheating spouse, give them a protonmail account. A divorce attorney might be able to tie cheatingwife44@protonmail.com to Mary Jones, but the trail will stop there. Unless Mary printed out her cheatingwife44s emails.
Great tip that’s realistic and useful for crime fiction authors. Thanks, Alan.
I hear you, Debbie. I turned off all AI features in Gmail a year ago to avoid this sort of thing. It did mean I lost the smart features such as sorting emails into primary, social, and promotion, but I’d actually rather just have one inbox.
Just the other day I was reminded of Proton mail as an alternative by another author who loves it for the privacy aspects.
AI is digital kudzu mated to digital ivy, it’s everywhere, but I’ll do my best to tear it out where I live, and keep it away when I can’t.
Dale, “digital kudzu mated to digital ivy” is the perfect analogy! I’ll be working beside you in the yard, tearing out the invasive noxious weeds.
Hi Debbie, I work in cybersecurity and everything you shared is an unfortunate reality. Even if you opt out or disable settings in various online platforms, they all (generally) reserve the right change their privacy policy “from time to time,” which could very well include them changing settings back that you had previously disabled. Every time I update my smart phone, I recheck all my privacy settings, because I have found settings that I’d previously disabled to be re-enabled after an update. So it’s recommended to check those settings every so often. It’s a never-ending battle that does not favor the privacy of the individual. Grrr…
Brock, welcome to TKZ! I’m always glad when a knowledgable tech person chimes in.
I too have run across mysterious resets after updates. B/c I’m not tech savvy, I always initially assume it’s my fault. Then I hear from more savvy friends who had the same update experience and I realize the fault wasn’t mine after all.
Debbie, it’s definitely not just you! 🙂
And I should clarify to my previous comment (so I don’t get sandblasted by any online platform compliance people!), I would say it’s uncommon for legitimate online businesses to change settings “back” that you have disabled. It’s more common practice (and less legally questionable) to introduce new “features” that may have new mechanisms for collecting information. I.e., your old settings remain in place, but the new feature settings may override what you intended to stop. Legit companies will disclose that information in the form of release notes, typically, but who reads those? 😉
Brock, do you mean that disclosure on page 475 in two point print?
Ha ha, exactly. =)
Debbie, Alan and Brock, thanks for this look at the new reality. I also get AI offers to summarize my manuscript. I’m tempted ask, but not after reading this.
Elaine, this feels like an end run around copyright protection. As if writers don’t have enough problems with piracy.
Elaine, I too have been tempted to upload my manuscript to an AI engine for edits and critique, but I just can’t pull the “trust” trigger yet despite the dislaimers (and even those with a configurable setting) to guarantee my work is not shared and assimilated into the AI Borg. I do paste in snippets of my WIP here and there for word usage/context analysis—no more than a paragraph—and of course AI always asks so eloquently, “Would you like me to rewrite this for you? I can make it sound SO MUCH better.” Not sure if I should laugh or get offended by that. Lol But I kindly respond, “No, thank you, just tell me if this word makes sense in this context and with the character I described to you, please.”
I’m sure I’ll get there, because I’m very interested in seeing what AI could do with an entire manuscript (again, just a copy edit and critique, please!), but then I have to wrestle with the question (and ensure I answer it honestly), “Did you use AI for this work? If yes, how much?” 🤓
I don’t use Gmail, but I wouldn’t be surprised if Outlook didn’t work the same way. They continuously ask me to use Co-pilot to rewrite my emails, which I’ve declined more times than I can count. Ugh. All these intrusive bots make me want to stay offline.
Sue, you and I can resort to handwritten letters to each other. 😉
Thanks for the warning, Debbie! What an intrusive world we live in. I assume everything I send in text or email will be read by someone it wasn’t intended for. On the other hand, I take heart that, given my lifestyle, I will personally be responsible for boring the bots to death.
We never send financial information through email. Our accountant has a secure site we can upload documents to.
Kay, I still write physical checks and insist on paper receipts for payments.
May you bore millions of bots into oblivion!
Another word that, in a different lifetime, might mean something totally different.
Bot…
This is disturbing info, Debbie.
OTOH, I just did a search on “How to kill a bot”. Hilarious-sounding links appeared, like “Chopinator”. For a good laugh in the face of things that go bump online, do your own search on it.
Bot on, Killzoners! 🙂
Deb, how to kill a bot sounds fun. Is that considered premeditated murder or a crime of passion?
Hmm…
🤡
This is scary to think about. How much info we share with friends and family as well as our writing.
Traci, thanks for stopping by.
I have a Gmail account that I never use, but must have for a lot of apps. I go over occasionally to delete all the ads that come in. As far as I know, keylogging apps are still outlawed on iOS, I’m sure it’s only a matter of time.
What worries me is that I’m sitting here listening to what has to be AI music, and it sounds good! I know it’s too perfect to be real, but AI is getting better and better.
Pat, even if AI replicates truly creative writing, I’ll keep writing b/c I love it.
Thanks. Just turned off every feature that has been annoying me – ALL of them.
It lost me the Categories – boo hoo – who cares?
Now I really have to figure out how to get my Mail back on my Mac. When I got a computer upgrade, it didn’t transfer properly, so I’ve been using Gmail as a ‘meanwhile’ – I hate all those features.
Alicia, good luck getting your mail program back. Those so-called features are filled with fish hooks.
I recall reading that “If you use a free service, you are the product.”
In my case I guess it’s a good thing that all of my personal information is pretty boring.
Queston of the day: Can a bot die of boredom?
Marshall, we are all “products” these days. Like cows being milked and milked and milked…