tech
Critics scoff after Microsoft warns AI feature can infect machines and pilfer data
Integration of Copilot Actions into Windows is off by default, but for how long?
TL;DR
- Microsoft's Copilot Actions, an experimental AI agent for Windows, can perform tasks but poses security risks.
- Known AI flaws like hallucinations and prompt injection can lead to data theft and malware installation.
- Microsoft advises experienced users to enable the feature only after understanding the security implications.
- Critics compare the warnings to those for macros, questioning their effectiveness in preventing exploitation.
- IT admins will have control over enabling/disabling Copilot Actions at account and device levels.
- Concerns exist about users' ability to detect AI exploitation and the eventual transition of experimental features to default.
- Microsoft's security goals for agentic features include non-repudiation, confidentiality, and user approval.