AI Model Access

Unified access to leading AI models through a single, secure interface

Provider Models Available Capabilities Context Window Status
Anthropic Claude 3.5 Sonnet, Claude 3 Opus, Claude 3 Haiku Chat, Analysis, Code, Documents 200K tokens Available
OpenAI GPT-4o, GPT-4 Turbo, GPT-3.5 Turbo Chat, Analysis, Code, Images 128K tokens Available
Google Gemini 1.5 Pro, Gemini 1.5 Flash Chat, Analysis, Code, Documents, Images 2M tokens Available
Perplexity Sonar Large, Sonar Small Research, Citations, Real-time data Varies Available
Meta Llama 3.1 405B, Llama 3.1 70B Chat, Code, Analysis 128K tokens Coming Q2

Image Generation

  • DALL-E 3 (OpenAI)
  • Gemini Imagen (Google)
  • Up to 4096×4096 resolution
  • PNG, JPEG, WebP output

Document Processing

  • PDF (text and scanned via OCR)
  • Microsoft Office (docx, xlsx, pptx)
  • Max file size: 50MB
  • Max pages per PDF: 500

Model Updates

New models added within 14 days of general availability. Deprecated models supported for minimum 90 days with migration assistance.

Data Protection

Automatic detection and filtering of sensitive information

Data Type Detection Method Action Options Accuracy
Social Security Numbers Pattern matching + checksum validation Block, Redact, Warn 99.9%
Credit Card Numbers Luhn algorithm + pattern recognition Block, Redact, Warn 99.9%
Email Addresses RFC 5322 pattern matching Block, Redact, Warn, Allow 99.5%
Phone Numbers International format detection Block, Redact, Warn, Allow 98%
API Keys & Secrets Entropy analysis + known patterns Block, Redact, Warn 97%
Custom Keywords Configurable word/phrase matching Block, Redact, Warn 100%

Processing

All filtering occurs before data leaves your control. Redacted content never reaches AI providers. Original content logged only if explicitly enabled by admin.

Custom Rules

Create organization-specific detection rules using regex patterns, keyword lists, or structured data formats. Unlimited custom rules on all plans.

Audit Logging

  • User ID and session info
  • Timestamp (UTC)
  • Model used, token count
  • Filter actions taken

Log Integrations

  • Splunk
  • Datadog
  • Sumo Logic
  • AWS CloudWatch
  • Custom webhook

Admin & Access Controls

Role-based permissions and authentication options

Role Capabilities Model Access Admin Functions
Owner Full platform access All models Billing, Users, Settings, Security
Admin Manage users and settings All models Users, Settings, Audit logs
Manager View team usage and reports Configurable Team reports only
Member Use AI tools Configurable None
Guest Limited access Restricted None

Authentication

  • Google Workspace SSO
  • Microsoft 365 SSO
  • Okta SAML 2.0
  • Custom SAML/OIDC
  • Email + password (optional)

Provisioning

  • SCIM 2.0 support
  • Auto-provisioning from IdP
  • Group sync
  • Just-in-time provisioning

Session Controls

  • Configurable session timeout
  • Force re-authentication
  • IP allowlisting
  • Device trust policies

Security Architecture

Enterprise-grade security controls and infrastructure

[***] Encryption

AES-256 encryption at rest. TLS 1.3 for all data in transit. Customer data encrypted with per-tenant keys.

[|=|] Data Isolation

Complete tenant isolation. Your data is never shared across accounts or used for model training. Data residency controls available.

{<>} Infrastructure

AWS US regions (us-east-1, us-west-2). Multi-AZ deployment for high availability. Auto-scaling infrastructure.

[//] Monitoring

24/7 security monitoring. 1-hour initial response SLA for security incidents. Public status page for transparency.

Network Security

  • WAF protection
  • DDoS mitigation
  • VPC isolation
  • IP allowlisting (Enterprise)

Access Security

  • MFA support
  • SSO enforcement
  • Session management
  • API key rotation

Data Handling

  • No training on customer data
  • Configurable retention
  • Data export capability
  • Right to deletion

Integration & API

Programmatic access to TrustML capabilities

API Features

  • RESTful API
  • OpenAI-compatible endpoints
  • Streaming responses
  • Batch processing
  • Webhook callbacks

SDKs Available

  • Python (official)
  • JavaScript/TypeScript (official)
  • Go (official)
  • Ruby (community)

Authentication

API key authentication with scoped permissions. Keys can be restricted by IP, model access, and expiration date.

System Requirements

What you need to run TrustML

[www] Browser

  • Chrome 90+ (recommended)
  • Firefox 88+
  • Safari 14+
  • Edge 90+
  • JavaScript enabled

[o-] Authentication

  • Google Workspace, or
  • Microsoft 365, or
  • SAML 2.0 IdP, or
  • Email for password auth

[~>] Network

  • HTTPS (port 443)
  • *.trustml.com access
  • WebSocket (optional)
  • No client install required