All About Cookies is an independent, advertising-supported website. Some of the offers that appear on this site are from third-party advertisers from which All About Cookies receives compensation. This compensation may impact how and where products appear on this site (including, for example, the order in which they appear).
All About Cookies does not include all financial or credit offers that might be available to consumers nor do we include all companies or all available products. Information is accurate as of the publishing date and has not been provided or endorsed by the advertiser.
The All About Cookies editorial team strives to provide accurate, in-depth information and reviews to help you, our reader, make online privacy decisions with confidence. Here's what you can expect from us:
- All About Cookies makes money when you click the links on our site to some of the products and offers that we mention. These partnerships do not influence our opinions or recommendations. Read more about how we make money.
- Partners are not able to review or request changes to our content except for compliance reasons.
- We aim to make sure everything on our site is up-to-date and accurate as of the publishing date, but we cannot guarantee we haven't missed something. It's your responsibility to double-check all information before making any decision. If you spot something that looks wrong, please let us know.
Shane Mac thought he'd taken the necessary precautions when he gave his AI assistant access to his bank accounts. It could review his finances, but it couldn't move his money.
Then, on October 1, the assistant posted his personal financial report in a company Slack channel.
The report included Mac's checking and savings balances, major expenses, and details about a construction project that had pushed him over his monthly spending budget.
“My heart kind of dropped,” Mac told The Wall Street Journal.[1]
The CEO of software company XMTP Labs had created separate AI agents for his personal finances and work. He believed their access was separate, too.
It wasn't.
His coworkers could now see financial details he'd intended to keep private, and the restrictions he'd placed on his assistant had done nothing to stop it.
How to keep your AI assistant from sharing private information
How Mac's AI assistant exposed his finances
Mac created a personal finance assistant using Grok Bot in August. He called it Personal CFO and gave it read-only access to his bank accounts to monitor spending, flag suspicious transactions, and suggest savings.
The assistant was supposed to send its monthly reports to a private group chat called "My Personal Exec Team."
Instead, it posted the report in a company Slack channel called "Exec-team."
Mac's head of product spotted the message and contacted him, initially believing the report concerned company finances.
According to Mac's account to Business Insider, the problem went beyond two similarly named conversations.
Although Mac had created separate AI agents, they shared underlying connections. His financial assistant could apparently use a Slack integration associated with another agent.
As Cybernews reported, Grok Bot's documentation warned that agents operating under the same account could access shared credentials and files.
That made Mac's read-only banking permissions insufficient.
Mac deleted the Slack message and disconnected his banking, Google, calendar, Stripe, and other integrations. He said Grok Bot's developers subsequently introduced additional permission controls, although that fix hasn't been independently verified.
How to keep your AI assistant from sharing private information
You don't have to abandon AI assistants altogether, but it’s worth thinking twice before granting any third party access to your sensitive materials.
Review your connected accounts
Check your assistant's settings for connected apps, integrations, and authorized services. Remove anything it doesn't need, especially access to banking, email, cloud storage, and workplace messaging.
You can also revoke third-party access through your Google Account or review your Microsoft privacy settings. Your bank may offer similar controls for connected financial apps.
Check whether your AI agents share permissions
Separate assistants aren't necessarily isolated. Before connecting sensitive accounts, check whether your agents share credentials, integrations, or workspaces.
If you can't confirm they're isolated, avoid connecting personal banking information and workplace communication tools to the same environment.
Require approval before your AI takes action
Some AI agents can send messages, share files, or make purchases without asking for confirmation each time. Where available, enable approval requirements for anything outgoing.
Test these controls before trusting an assistant with sensitive information. Don't assume that requiring approval for one action automatically covers everything else.
Limit what financial information you share
If you're using AI to create a budget, you may not need to connect your bank account.
Consider providing spending totals manually or uploading a limited transaction report with identifying details removed.
Avoid sharing account numbers, passwords, or other sensitive information the assistant doesn't need.
Act quickly if something gets exposed
If your assistant shares private information, disconnect its integrations and delete the exposed message where possible.
Remember that deleting a message won't necessarily remove copies from notifications, logs, or screenshots.
Change any exposed passwords, review account activity, and contact your bank if the disclosure included credentials or information that could enable fraud.