Russian Police Warn AI Agents Need Careful Setup to Protect User Data
AI agents that work on their own with files, browsers, email and online services can open up extra security holes if they are configured carelessly, Russia's Interior Ministry has warned.
The ministry's unit handling the fight against criminal use of information and communication technologies said such assistants must be set up safely before they are put to work.
Its advice: let an agent reach only selected folders rather than an entire disk, and set a list of approved websites so it cannot wander onto unknown resources or send them data.
Users should also require confirmation before an agent sends emails or deletes files, and purchases and payments should need the user's consent. Unneeded plugins, integrations and command-line access should be switched off, while active permissions and access tokens should be reviewed and those no longer needed revoked.
The ministry's specialists also recommend extra isolation tools that run AI agents in a separate virtual environment, limiting their direct access to the user's system and accounts.
The safety of an AI agent, the police noted, depends not only on the neural network itself but on how tightly its powers are limited.
Source: RIA Novosti
dArt Studio installs AI for local businesses in Broward & Palm Beach County, FL. We reply within 1 business hour.