Discover Top American-Made Products!
Support local craftsmanship with these high-quality, American-made items—shop now on Amazon!
Shop NowYour AI productivity tools may be blabbing trade secrets. High-tech office gadgets save a lot time and effort. At the same time, users need to be really careful not to let a robot bypass corporate security. Big Brother could easily build in back doors and hackers manage to break into everything connected to the internet, sooner or later. Audio clips of office background chatter might not seem like much but could be worth a fortune. In this case, it wasn’t hijacked or used intentionally for foul play but uncovered information not meant for the person who received it.
Secrets and AI gossip
Otter AI has the potential to leak “company secrets and workplace gossip.” Tech researcher and engineer Alex Bilzerian, discovered an “innovative office tech that utilizes artificial intelligence” to do some tedious transcription for him. That’s how he learned about an exploitable flaw.
In order to do the intended task, like transcribe a Zoom meeting, the robot must be “listening in on your conversations.” It can keep working long after you though it was done, “turning the tool into a dangerous nuisance.”
That’s how Bilzerian caught on. The robot “accidentally spilled a confidential conversation.” He asked it to type up a transcript after a “Zoom meeting with a venture capitalist firm.” He knows now, after looking a lot closer at the documentation, that “Otter AI prioritizes the privacy of its users.” That’s why “users can manage who receives a copy of meeting transcripts in their account settings.”
Bilzerian’s got circulated around by email to everyone in the meeting. When he opened his, he got a major surprise. It was full of his investor’s secrets. Apparently, “the smart assistant had continued to record the conversation even after Bilzerian had logged off.” What they said behind his back was a dealbreaker.
He thought the emailed transcript would give him a record of his part of the meeting for his notes. To his amazement, it went way beyond that. It also included “hours of their private conversations afterward, where they discussed intimate, confidential details about their business.”
In this case, “their” being investors from the venture capital firm. They didn’t have nice things to say about him. “While the investors ‘profusely apologized,‘ Bilzerian still decided to stop the deal with their firm.” Then he realized nobody’s secrets are safe.
Not a reasonable assumption
It “might be a ‘reasonable assumption‘ to think the AI assistants could detect when participants exit a meeting and not send the rest of the transcript.” Nobody programmed Otter to figure that out.
Hatim Rahman, an associate professor at the Kellogg School of Management at Northwestern University points out “the tech is not always accurate.” If you have secrets, be careful about talking about them where a microphone could be listening.
Otter AI responded to Bilzerian’s social media post. They explained that they understand his concerns and are “committed to keeping your information private and secure.” That’s not helping the investors who’s secrets got blabbed. The ones victimized weren’t even aware the software was in use. The solution they offer wouldn’t have been available to the victims.
“For this specific instance, users have the option not to share transcripts automatically with anyone or to auto-share conversations only with users who share the same Workspace domain.”
Next time, Bilzerian will set it so only he gets a copy then sit back and wait to see what corporate dirty secrets he can harvest without the company’s knowledge. The good news is that OtterPilot “will not record anything said by a muted participant.”
Even though it will record hours of background chatter from all the other microphones. “The company can also collect screenshots of meetings — including text or other media uploaded by participants.“
