Skip to content
CollaborateGet updates

Responsibility

Safeguards and usage policies

What each model and product will not be used for, and the work we decline.

Usage policies

Life-sciences models (truth-hdi, truth-ls)

  • No diagnostic, prescribing or treatment claims. Outputs support qualified professionals.
  • Not offered to patients as medical advice.
  • Every output carries its evidence and abstains when evidence is insufficient.

Realtime video and Media Studio

  • No generated likeness of a real person without their consent.
  • No content designed to deceive viewers about what is real.
  • Generated media carries provenance information.

Media intelligence and Media Monitor

  • Tracks organisations and public topics, not private individuals.
  • Information about journalists is limited to their published work.
  • Not offered for surveillance of individuals or groups.

Software from intent and Viber

  • Generated apps are checked for common data-handling and access-control failures.
  • Not for building software intended to harm, deceive or surveil people.

Work we decline

  • Surveillance of individuals or groups.
  • Political persuasion or targeted messaging during elections.
  • Synthetic media of real people without consent.
  • Medical products that make claims without clinical evidence.
  • Training on data we do not have the right to use.

We count what we decline and publish the tally in the annual report, without naming the party that asked.

Before release

Every release is tested for misuse before it ships. A summary of that testing, and what changed as a result, is published with the release.