The Lost Cause of Locking Down AI in the Workplace

Credit: Mohamed Nohassi/Unsplash

Last week was a huge week for AI. Moreover, it was a huge week for consumer AI. Between ChatGPT’s new interrupt-me-anytime-you-want voice interface and Google’s chatty Project Astra that can see everything you point your phone at, it really feels like we’re about to enter an era where a true AI assistant is just a tap or a wake word away.

In the wake of the announcements, there was understandably a lot of focus on who one-upped who, and which company scored more points with its demos. If you zoom out, though, you can see both companies pushing toward something: a new milestone in the race to build intelligent machines — one where the consumer interface gets so convenient that it becomes difficult to not use AI, since it will be powerful, seamless, and embedded into everything.

Just think: Google’s making AI search — where it basically just gives you the answer before it shows you a page of links — the default. ChatGPT will have real, verbal conversations with you while you show it your desktop via the Mac app. Or your real desktop, via your phone camera. If you connect ChatGPT to your iPhone’s action button, you won’t even need to think about using the app. What a world.

That picture has lots of appeal, but it also has a problem: most businesses aren’t ready for that world.

Subscribe now

The Privacy Question

The No. 1 thing I’m asked about whenever I hold a training class for a marketing or PR company is, “What about privacy?” Certainly, any company that deals with client information (read: the vast majority) urgently needs to know what they can or can’t share with an AI. What these chatbots or apps will do with the data they give it is vital when you’re dealing with embargoed or confidential info.

And the answer, at least with consumer-level AI, is that it’s complicated. Most public-facing AI services will harvest the information you feed into these chatbots as training data. And although the chance of that data being “regurgitated” when someone else puts in a random prompt is extremely tiny, it’s not zero. This possibility is what leads many companies to forbid the use of public chatbots entirely.

Yes, any company can easily get around this issue by restricting their use of AI models to APIs, or services built around APIs, which explicitly don’t harvest training data. And many do. However, for small-to-medium-size businesses, this becomes a huge headache: It requires either dealing with an array of vendors for AI applications, each with their own fee structure and privacy policies, or investing a lot more in product and engineering to build your own.

Even worse, using AI in this way cuts you off from the shiny new thing — the AI features that are being built into the consumer apps and devices. Not that the companies care, but their employees do. And that could end up being the biggest force of all at play here: In much the same way the iPhone made “bring your own device” a thing in the workplace, this next wave of AI experiences could fuel a “bring your own AI” movement. Your enterprise AI portal might be able to do some cool stuff with spreadsheets, but can you chat with it about the sushi you’re making?

Leaks From Outside

This is the danger that all those businesses who are concerned about data privacy are facing: more and more of their workforce adopting consumer AI and inevitably bringing that to their jobs. You might be able to train them to use the company’s AI one way and their personal AI another, but the more they use ChatGPT or Gemini (or whichever Apple puts on the iPhone), leaks between the two will inevitably happen.

This puts pressure on businesses to implement strict policies around consumer-facing AI. But it also creates an opportunity: Although there are plenty of businesses vying to be the leader in “enterprise” AI, no one is near declaring victory. The demand for private, “safe” AI is very real, and those trying to meet that demand shouldn’t overlook the rapid evolution of consumer experiences.

It’s easy to dismiss consumer-centric features like ChatGPT’s new voice mode as just bells and whistles. But BlackBerry used to say the same thing about the iPhone. Look where that got them.

The Media Copilot is a reader-supported publication. To receive new posts and support our work, consider becoming a free or paid subscriber.

Ready to start using AI like a pro?


Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.