What is one specific design choice that makes AI agents better protect user privacy?

The answers all emphasize giving users more control and transparency over data access, permissions, and actions, though they differ in focusing on documentation, approval, memory, or technical restrictions.

5 answers

Answers

  • Dean Ball

    Head of Strategic Futures · Openai

    Limited, Intelligible Permissions

    Dean Ball says AI agents should be built with permissions that are intelligible and limited, alongside auditable actions and withdrawable authority, as a way to make commitments to user privacy.

    Don't Worry About the Vase · Don't Worry About the Vase · Read
  • Timnit Gebru

    Founder & Executive Director · The Distributed AI Research Institute (DAIR)

    Data Documentation

    Timnit Gebru says companies should document where their data came from before releasing a product, calling data documentation a simple transparency step they currently avoid.

    The answer starts at 13:44.

    Uncanny Valley | WIRED · Listen on Apple Podcasts
  • Reid Hoffman

    Co-Founder, Executive Board Chair · Manas Ai

    Human Approval Before Sending

    Reid Hoffman says he tells AI agents to read communications and draft replies but not send anything until he approves it, describing this as a practice he adopted after learning it the hard way.

    Reid Hoffman · Watch on YouTube
  • Ina Fried

    Chief Technology Correspondent · Axios

    Selective Memory

    Ina Fried says the design choice is selective memory, letting users tell a service what to remember and what not to, because she wants to be in control while those controls stay manageable for an average busy person.

    Pioneers of AI · Watch on YouTube
  • Rob Allen

    Chief Product Officer · Threatlocker

    Restrict File Access

    Rob Allen says he set up Claude so it could only access files in a folder called "claude," arguing that restricting access this way is a sensible approach for letting people use AI tools without exposing sensitive information.

    Security Weekly - A CRA Resource · Watch on YouTube

More Generative AI questions

View all →