Snapshot 192936

Document
Security page
URL
https://www.aithon.ai/security
Fetched
HTTP status
200
Content type
text/html; charset=utf-8
Fetch mode
static
Size
71993 bytes
SHA-256 (raw)
732f37101e87439ca0eaa1b805155221d71e6d5ddf51eccf2900bac8608e466b
SHA-256 (normalized text)
05192d596817d843b3abb9ebd937ce1f6aa01609001e43834f61129b325ec49f

Normalized text

Scripts and page chrome removed; this is what change detection compares.

SOC 2 TYPE II
Security built for the way regulated fintechs actually buy software.
Aithon processes your sales activity — pipeline, communications, meetings — to make your reps smarter. We never touch your end-customer banking data. CRM and email writes happen only as part of campaigns and sequences your reps explicitly run, with scoped OAuth, full audit trails, and admin-controlled enable/disable per integration.
Talk to securityRequest architecture deep-dive
POSTURE · AT A GLANCE
A short list of what's already true on day one.
SOC 2 Type IICertified
DPAStandard DPA available
EncryptionTLS 1.3 · AES-256 · managed keys
AuthOAuth 2.0 · OIDC · MFA-ready
InheritedCloud provider SOC 2 · ISO 27001 · CSA STAR
01 · Integrations & auth
Every connection scoped, admin-consented, revocable.
Aithon authenticates to every source system through OAuth 2.0 with explicit admin consent. Read scopes power intelligence; write scopes are separate, optional, and only used when your reps run campaigns or sequences. No service accounts, no stored passwords, no standing credentials.
Source	Auth	Access scope	Mode
Salesforce
OAuth 2.0	Read: accounts, opportunities, contacts, pipeline.
Write (campaigns): when a rep runs a sequence, Aithon updates contact engagement, logs activities, and adds rep-approved notes — using a separate, admin-consented write scope. Every write is logged and reviewable in the activity audit. Disable per integration at any time.	Read · Write (campaigns)
Outlook
OAuth 2.0	Read:sales mailbox content & metadata for relationship signals.
Write (sequences):when a rep runs a sequence, Aithon sends emails on the rep's behalf using their delegated permission. Drafts can be set to require rep approval before send or auto-send per sequence policy. Every send is logged with the originating sequence ID.	Read · Write (sequences)
Microsoft Teams
OAuth 2.0	Meeting transcripts & metadata via Graph API OnlineMeetingTranscript.Read.All. Per-user delegated permissions; transcripts processed only when the user has consented to the integration.	Read-only
Highspot
OAuth 2.0	Content engagement analytics (which deck was viewed, by whom, when), titles, categories, and the document content itself for retrieval and grounding. Used to power content recommendations and ground AI responses in your enablement library.	Read-only
No stored passwords.
All access is token-based. OAuth tokens are stored Fernet-encrypted in Postgres, auto-refresh themselves, and never appear in logs or env vars.
Writes are opt-in, per integration.
Read and write scopes are granted separately. An admin can enable read-only intelligence first, then enable write scopes for campaigns and sequences once they’re ready. Either can be disabled at any time without losing the other.
Full write audit trail.
Every Aithon-initiated CRM update or email send is logged with timestamp, user, integration, originating sequence, and payload diff. Available to your admins in the activity audit and exportable as CSV.
02 · Privacy controls
Your admins switch off any data category at the source.
Most “privacy controls” in B2B SaaS are post-processing filters. Aithon's are a hard gate at the pipeline boundary — when your admin marks a category Excluded, that data is never loaded into the AI processing layer in the first place.
Hard gateEnforcement at the gold→insights boundary, not a post-filter on results. Excluded data never enters the AI pipeline.
RetroactiveFlipping a category from Eligible to Excluded purges previously-derived insights from the insights layer. Originals in bronze/silver/gold are untouched.
Fail-closedIf the privacy-control service is unreachable, the pipeline defaults to PRIVACY_FAIL_MODE = closed. Excluded by default rather than processed without authorization.
AuditableEvery change logged with timestamp, admin user, category, and old/new state. Visible to your admins in the Settings dashboard.
aithon · admin · /privacy-controls
AI Data Eligibility
Choose which data categories the AI processing pipeline is permitted to use. Excluded data is never loaded into AI; existing insights for excluded categories are purged within 24h.
Email CommunicationsOutlook · 24,841 messages indexed
On
Call TranscriptsGong · 412 calls last 30d
On
CRM RecordsSalesforce · accounts & opps
On
Meeting NotesTeams transcripts · Granola
On
MessagingSlack channels
Off
Stakeholder ProfilesContact enrichment
On
ENFORCED AT PIPELINE BOUNDARY · LAST UPDATED 2 MIN AGO BY YOU
03 · Defense in depth
Identity, network, data, application — controls at every layer.
No single control is sufficient on its own. Aithon implements specific guardrails at each layer of the stack so that a compromise at one level does not cascade to the next.
Identity & access
OIDC · JWT · SSO
Org-scoped auth · multi-tenant isolation
JWT tokens with role claims (admin / user)
All API requests validated server-side
No passwords stored — delegated to identity provider
MFA configurable per organization
Network
Private network isolation
All services in private subnets
Load balancer is the only public-facing endpoint
Security groups enforce least-privilege ports
Database & cache not publicly accessible
Private DNS for service-to-service
Data
Encryption everywhere
In transit: TLS 1.3 (load balancer · service-to-service)
At rest: AES-256 — database · object storage · block volumes (managed keys)
Secrets in dedicated secrets manager (no env vars)
OAuth tokens encrypted at the application layer
Backups encrypted · 7-day retention
Application
Secure by design
Org-scoped queries · no cross-tenant access
Server-side rendering · no client API keys
Parameterized queries · no raw SQL
Input validation on every endpoint
CORS restricted to known domains
04 · Deployment models
Multi-tenant by default. Dedicated infrastructure when you need it.
Most customers run on shared infrastructure with logical isolation at every layer. Enterprise customers with stricter requirements can be deployed to a dedicated compute cluster, database instance, object storage bucket, and customer-managed encryption keys.
● Standard
Multi-tenant
Shared infrastructure with logical isolation. The default for the vast majority of customers.
Shared compute cluster
APIWebAppAgents
Shared database
Org AOrg BYour orgOrg D
Org-scoped queries — aithon_org_id on every table
Identity-provider organization per customer — JWT-enforced
Per-org privacy controls & pipeline config
Encrypted credentials per-user
No API endpoint returns cross-org data
● Enterprise option
Dedicated infrastructure
Isolated compute and storage for the strictest security postures. Available as an enterprise option.
Dedicated compute cluster
APIWebAppAgents
Dedicated database
Your org · only tenant
Separate compute cluster & task definitions
Dedicated database instance
Isolated object storage bucket for the data lake
Separate private network or dedicated subnets
Customer-managed encryption keys (optional)
Independent scaling & maintenance windows
Ready for your security team to take a look?
For a deeper review — integration scopes, threat model, write audit walkthrough, regional residency — book an architecture deep-dive with our team. We'll bring an engineer.
Talk to securityBook architecture deep-dive