AI architecture and public-source research

Market map

See where AI is used, which security approaches may apply, and which models, tools, business systems, and data AI can reach. Explore solution approaches and public-source research for 66 currently displayed vendors.

Enterprise AI architecture

AI activity, security approaches, and connected systems

Read from where AI is used, through the security approaches connected to that use, to the models, tools, business systems, and data AI can reach. Select an approach to open its details beside the map.

Choose a use caseFull architecture

Choose one use case to show the security approaches connected to it.

Full architectureShow every security approachHow do we protect employee use of AI assistants and sensitive data?Protect employee access to AI assistants and prevent sensitive prompts, files, and responses from moving where they should not.How do we secure AI applications and action-taking agents?Protect AI applications and agents as they use models, connect to tools and business systems, handle credentials, and take actions.How do we govern AI risk, policy, and compliance?Create an inventory of AI systems, assign owners, set policies, manage risk reviews, and retain evidence for audits.How do we test and assure AI systems?Test how AI systems fail or can be misused, review model and software artifacts, document findings, and verify remediation.How do we attribute and control AI spend?Track who and what is consuming AI services, set budgets and usage limits, and detect unexpected spend.

Current viewShowing the full architecture

Showing all security approaches in the research model.

Stage 1
Where AI is used
- Workforce AI use
  Employees use assistants, browsers, and embedded AI features in daily work.
- Enterprise assistants and approved AI platforms
  Sanctioned enterprise assistants and AI workspaces.
- Custom AI applications and knowledge retrieval
  Internal copilots, model-backed applications, and systems that retrieve company knowledge.
- Agents and coding systems
  Action-taking agents, coding assistants, developer tools, and automated workflows.
Stage 2
Security approaches
- Agent and machine identity
- Employee AI access and data
- Application and agent safeguards
- Governance, testing, and visibility
- Usage and cost controls
Stage 3
What AI can reach
- Models and providers
  Hosted, internal, fine-tuned, and embedded models.
- APIs, AI tools, and connectors
  Services that retrieve information or perform actions.
- Enterprise applications and business systems
  Systems where AI reads or changes business records.
- Sensitive data, retrieval and memory
  Databases, files, vector stores, prompts, and memory.

Market map

AI activity, security approaches, and connected systems

Enterprise AI activity and security approach architecture