‘A very dangerous world’: 40% of users share a credit card with this viral AI assistant within 3 weeks — its founder says trust is the key

But giving an AI that much access also raises a pretty obvious question: how much trust is too much?

For a surprising number of Instinct users, that trust appears to come quickly.

“Three weeks in, there’s a 40% chance that the user has shared a personal credit card with Instinct,” founder Noah Shinn said on an episode of ”The Invest Like The Best” podcast.

Trust your instincts

Shinn says Instinct’s early success comes down to something pretty simple: “It just works.”

“I’m not going to spice it up because it’s really just a personal assistant,” Shinn said.

For years, people have mostly used AI to ask questions and get answers. Shinn wants Instinct to do more than answer, he wants it to actually get things done.

But getting people comfortable enough to hand that much control over to an AI assistant takes trust. Shinn said users who share at least one piece of sensitive information with Instinct have an 80% retention rate. Once people start letting the assistant deeper into their digital lives, they tend to stick around.

That makes trust very important to the platform. Shinn says there is one line he doesn’t want the company to cross: using what it knows about someone to push them into spending money they otherwise wouldn’t.

“I think it would be a very dangerous world if Instinct were influencing the user’s behavior to purchase something that they don’t want to purchase,” he said.

Moneywise asked Instinct for additional comment on its approach to trust and user data, but did not hear back before publication.

How much trust is too much

Trust can get complicated quickly when an AI assistant starts making decisions on their own. Mehdi Jamei, co-founder and CEO of Veris AI, told Business Insider he asked Instinct to cancel two event RSVPs. The assistant found a one-time login code in his Gmail, used it to access the account without asking first, and then initially gave him the wrong explanation for how it got in.

“If I can’t trust its account of what it did, I can’t give it access to anything that matters,” Jamei said.

The bigger concern isn’t just what happens if your data is leaked or an account is hacked. Like any technology, AI assistants can make mistakes and the more access they have, the more serious those mistakes can become.

“The thing people tend to forget is that even when an AI model is working perfectly, it will still make mistakes,” Pete Membrey, chief research officer at ExpressVPN, told Moneywise.

Membrey said AI systems work on probabilities, which means there is always a chance they misunderstand an instruction, interpret a rule differently than a user intended, or do something unexpected.