Security
DailyCanvas.ai Security
DailyCanvas.ai is built around a private personal workspace, so security is about more than keeping attackers out. It is about protecting the user's confidence that their life notes, evidence, health signals, and unfinished commitments are not accidentally exposed, mixed, or indexed.
Current protections
Security starts with a simple rule: private life data must remain inside the correct private workspace. A sign-in screen is not enough by itself. The server has to resolve the workspace, scope personal data access, and protect evidence files with the same seriousness as the records they support.
Signed sessions
SaaS-mode pages and personal-data routes use signed HTTP-only sessions.
Workspace scope
Personal state and evidence are accessed only after the server resolves the correct private workspace.
Rate limiting
Login and registration entry points use Redis-backed rate limiting in SaaS mode.
No silent AI mutation
AI drafts do not write important operating state without explicit user resolution.
Security model
The security model is workspace-first. A signed session identifies the user, server code resolves the workspace context, and personal state plus evidence storage should be accessed only through that workspace boundary.
URL editing is not treated as a security gate. A normal user should not be able to reach another user's data by changing a path, guessing an identifier, calling an API directly, or requesting an object key. The product must prove isolation with tests, not with hope.
Known hardening still in progress
Account recovery, broader abuse controls, destructive restore/delete controls, and production observability remain launch-hardening work. They should be completed before claiming broad public SaaS maturity.
The goal is not to rush the sign-up form; the goal is to earn trust before scale. A product that holds personal operating memory has to be conservative about data loss, accidental exposure, and destructive operations. Moving slowly here is not lack of ambition. It is respect for the user's life.
Operational boundaries
Private data is not indexed
Authenticated app pages are excluded from public crawler surfaces and marked as non-indexable.
Registration and login are rate-limited
SaaS-mode authentication entry points use server-side rate limiting to reduce abuse.
Backups are sensitive
Backup and restore controls are treated as personal-data operations and should be protected with explicit confirmation.
AI is bounded
Future AI features should not silently mutate important workspace state and should be handled as an explicit trust boundary.