What the AI reads, and what we keep
The privacy notice says what an audit's AI scan reads, what we keep and when the list of flagged items is deleted. The AI engineering page says what happens to your email during a build.
Fixes and improvements
- An audit's scan runs on Claude, or Cohere if you ask, with model training turned off; a person reads only what it flags
- We keep the roadmap and the flagged list, store no copies of your emails, and delete the list 30 days after delivery
- During a build, our access ends at handover; for support, you give us access again for each fix
- The data promise reads: we use paid AI accounts with model training turned off