SOLVED: Why You Might Want To Disable Generative AI Systems

Unless you’ve been living under a rock you will be familiar with generative AI and if you’re a tech or a CTO, you should already know that there are a number of privacy issues with generative AI.

We have a simple article explaining how to disable Adobe Acrobat AI features via the GUI, the registry, or Group Policy you might find helpful.

The biggest issue is that generative AI systems like ChatGPT and Adobe AI will use the text or graphics from users questions, in their future answers. Other words, Adobe AI, ChatGPT and others ingest what you ask and make it part of their model, which means it can easily expose confidential information.

Take these three examples a tech or a staffer might ask of ChatGPT or Microsoft Copilot:

• Take this Visio document showing all of our network connections and put it into a PowerPoint that non technical people can understand
  • That Visio file showing all of your network is now part of whatever AI system you were working with. Hackers will love you for it
• Use this payroll Excel file to provide percentile breakdown by sex and age
  • Your payroll file is now public because the AI system you’re using just ingested it. Not funny.
• Write a script to change the users in Elections Canada’s Organizational Unit named “confidential”
  • If this gets out, hackers will know that Elections Canada has an Organizational Unit named “confidential” in its domain

Many AI companies, like Microsoft, have put policies in place to protect corporate data from being ingested. Usually the free services ingest Everything you ask or reference but if you are paying for a service like Microsoft Copilot you can opt out of that and stop your information from being added to the giant AI model.

However there are a few companies that have stubbornly refused to stop their AI’s for using every bit of data and every question used by it. The most notable of these AI systems comes from Adobe. Adobe has stated, whatever you put into their generative artificial intelligence systems becomes owned by them and ingested into their AI models and can easily be exposed to others.