Forever online

Editor's note: The following text is a transcript of a podcast story. To listen to the story, click on the arrow beneath the headline above.

MARY REICHARD, HOST: It’s Thursday the 19th of February.

So glad to have you along for today’s edition of The World and Everything in It. Good morning, I’m Mary Reichard.

NICK EICHER, HOST: And I’m Nick Eicher.

First up: artificial intelligence and its long memory.

The power and popularity of AI has skyrocketed in recent months, and as these tools become more a part of daily life, the information you share may last longer than you think.

REICHARD: WORLD’s Mary Muncy reports.

MARY MUNCY: This month, Amazon’s home security service Ring expanded its ability to search for lost dogs to most users in the U.S. The feature, called "Search Party," uses AI to track lost dogs. The company announced it in a Super Bowl ad.

RING COMMERCIAL: This is Milo. / Pets are family, but every year, 10 million go missing, and the way we look for them hasn’t changed in years...until now. One post of a dog’s photo in the Ring App starts outdoor cameras looking for a match.

The backlash on social media was swift and full of references to George Orwell’s 1984, questioning whether the feature would really only be used to track dogs.

INSTAGRAM REEL: I know what you’re thinking, no, it’s not the same thing as the book. This one’s only for dogs, because we said it was.

WYZE COMMERCIAL: But what if we could make finding one lost dog require the computational power of a small, dictator-led, nation-state?

The company also had plans to streamline a feature that allows police to request video footage from Ring users by partnering with Flock, a company that makes license plate readers. But this week, the companies canceled that partnership. 

While Ring still has a limited ability to share video with law enforcement and the dog search feature is still active, both can be disabled.

ChatGPT has also faced recent criticism over data privacy. This month, the model’s owner, OpenAI, began testing targeted ads. The company says in free and go accounts, ads for items or services related to a user’s conversation with ChatGPT will start appearing at the bottom of an answer.

Competitor Anthropic parodied that in an ad that begins with a man asking for a workout plan for a six-pack.

ANTHROPIC AD, CHATGPT: Perfect, let me personalize this for you. Let’s start with your age weight and height. Whenever you’re ready.

GUY: 5’7” 23-years-old, 140 pounds.

CHATGPT: Got it, I’ll create a plan that focuses on aesthetic strength training. But confidence isn’t just built in the gym. Try Step Boost Maxx. The insoles that add one vertical inch of height and help short kings stand tall.

OpenAI’s chief ad officer Asad Awan says they have some safeguards to prevent that sort of scenario.

OPENAI: Number one, the answers need to be independent from the ads, both visually, but also in how the models are trained and how the system works so that you can always trust the answer.

Awan says OpenAI tailors the ad to the conversation, and that advertisers themselves have no access to users' conversations.

So just what can AI companies do with data that you give them? And how can you protect yourself as new technologies emerge?

Todd Jacobs is the executive director of the Theia Institute, a think tank studying AI. He says the small piece of information that you give one website here or ask ChatGPT about there are likely harmless... individually.

JACOBS: It's the ability to tie these seemingly unconnected pieces of information together across the whole spectrum of your life that is invasive of your privacy.

For example, say you bought a book on starting a business, and then you bought an all-in-one kit for carpentry online, then you start seeing ads about small business loans. 

That means the book and carpentry websites likely sold your data to another company, and that company sold it to their partners, and so on.

JACOBS: On some level, that is something that could be good for people, but it's done in a way that ensures that you actually have almost no control over your data.

The same goes for large language models like ChatGPT. Once you put information in there, that company owns it—whether that’s your phone number and address at the top of your resume, or the family photo that you want to clean up.

That may not be a big problem now, but…

STAN MATWIN: Today it's owned by Amazon, but maybe in five years, Amazon is bought by Nvidia.

Stan Matwin is a professor emeritus of computer science at Dalhousie University.

MATWIN: The ownership of the data may change, and it may go from an organization that I like, Amazon, I believe, uses it responsibly, to somebody who's much less responsible and will exploit every aspect of it.

There are settings on just about every platform that tell a company how to use your data. But generally, the default settings give the most information and freedom to the company, and that information generally doesn’t go away without a request to delete it.

You can check what information is out there about you using a tool like the Wayback Machine, which stores internet pages in a static form even if its been deleted elsewhere.

JACOBS: It is ultimately accessible and—barring their data center being hit by a meteor—largely forever.

But ultimately, Jacobs says most people need to be cautious, not paranoid.

JACOBS: We think about this in terms of trade offs, right? What, what are you willing to give up in exchange for additional privacy?

The answer may be using a more secure browser that doesn’t run your favorite video game. Or not downloading the latest app because it asks for permission to things it has no business being in… like a photo app asking for access to your music library.

Jacobs says the best way to protect your privacy is by not putting sensitive information on the internet anywhere—especially in a large language model.

The Federal Trade Commission considers quietly changing a privacy policy to be a deceptive trade practice, meaning generally companies will notify you… so, Jacobs says make sure you check the right boxes and skim privacy permissions the first time…. And every time you’re prompted to.

Reporting for WORLD, I’m Mary Muncy.

WORLD Radio transcripts are created on a rush deadline. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of WORLD Radio programming is the audio record.