NEW · How to verify an AI-written citation in four minutes — read it →
Risk

Running a local model for privileged work

Oct 19 2026 · Risk

A local model keeps privileged material on your machine. Here is the honest trade-off and a pragmatic setup.

Short answer

A local model runs on your own hardware, so nothing leaves your network — the strongest answer to privilege concerns. The trade-offs are real: local models are less capable than frontier cloud models, need a decent machine, and take effort to set up. A pragmatic approach is to use cloud models for public and confidential material with training disabled, and reserve a local model for the genuinely privileged work where the confidentiality guarantee matters most.

The trade-off, stated plainly

  • Pro: nothing leaves your network.
  • Con: less capable than the best cloud models.
  • Con: needs hardware and setup time.

A pragmatic split

Most work is not privileged at the point of drafting. Use the cloud for speed, and switch to local for the small set of matters where confidentiality is the whole point.

Get the confidentiality pack

Written for legal professionals. Not a guarantee of accuracy — always verify every claim against its source, and let a named human sign off. Get the free handbook.

Running a local model for privileged work · LexGPT