Five TSC categories. Security is mandatory ("Common Criteria"); availability, processing integrity, confidentiality, privacy are optional based on service org choice. Each TSC has criteria and points of focus drawn from COSO Internal Control — Integrated Framework.
Security (Common Criteria, CC)
Mandatory. Nine CC sub-categories:
CC1: Control Environment
- CC1.1: integrity and ethical values
- CC1.2: board oversight
- CC1.3: organizational structures
- CC1.4: commitment to competence
- CC1.5: accountability
CC2: Communication and Information
- CC2.1: relevant information about objectives
- CC2.2: internal communication of information
- CC2.3: external communication of information
CC3: Risk Assessment
- CC3.1: objectives specified
- CC3.2: risks identified
- CC3.3: fraud risk assessment
- CC3.4: significant change identification
CC4: Monitoring Activities
- CC4.1: ongoing and separate evaluations
- CC4.2: evaluating and communicating deficiencies
CC5: Control Activities
- CC5.1: control activities selected and developed
- CC5.2: technology controls
- CC5.3: policies and procedures
CC6: Logical and Physical Access Controls
- CC6.1: logical access
- CC6.2: account management
- CC6.3: removal of access
- CC6.4: physical access
- CC6.5: physical access removal
- CC6.6: logical access protection
- CC6.7: data transmission
- CC6.8: protection against malicious software
CC7: System Operations
- CC7.1: vulnerability detection
- CC7.2: anomaly monitoring
- CC7.3: incident response
- CC7.4: incident recovery
- CC7.5: recovery from outages
CC8: Change Management
- CC8.1: change management process
CC9: Risk Mitigation
- CC9.1: vendor / business partner risk
- CC9.2: vendor / business partner management
Availability
Optional TSC. Criteria:
- A1.1: capacity / system performance
- A1.2: environmental protections
- A1.3: backup / restore
Processing Integrity
Optional. System processing is complete, valid, accurate, timely, authorized.
- PI1.1: definitions and policies
- PI1.2: input controls
- PI1.3: processing controls
- PI1.4: output controls
- PI1.5: storage controls
Confidentiality
Optional. Information designated confidential protected.
- C1.1: identification and protection of confidential information
- C1.2: destruction of confidential information
Privacy
Optional. Personal information collected, used, retained, disclosed, disposed per commitments.
- P1.0: management policies for privacy
- P2.0: notice
- P3.0: choice and consent
- P4.0: collection
- P5.0: use, retention, disposal
- P6.0: access (data subject rights)
- P7.0: disclosure to third parties
- P8.0: security for privacy
- P9.0: quality
- P10.0: monitoring and enforcement
GDPR overlap substantial; Privacy TSC partially aligns with GDPR principles.
Selecting TSC categories
Service org chooses which optional TSCs to include in scope:
- Security only — minimum scope.
- Security + Availability — common for SaaS.
- Security + Availability + Confidentiality — common for SaaS handling customer business data.
- Security + Availability + Confidentiality + Privacy — common when processing personal data.
- All five — comprehensive.
Customer-procurement preferences drive selection.
Points of focus
Each TSC sub-category has "points of focus" — illustrative considerations to assess whether the criterion is met. Not mandatory controls; org designs controls addressing relevant points of focus.
SOC 2 control activities
Service org designs control activities mapping to TSC criteria. Control activities are specific operational practices:
- Access reviews (CC6.2, CC6.3)
- Vulnerability scanning (CC7.1)
- Incident response (CC7.3, CC7.4)
- Change management (CC8.1)
- Vendor management (CC9.1, CC9.2)
- Encryption (CC6.7)
- Logging and monitoring (CC4.1, CC7.2)
- Training (CC2.2, CC4.2)
Auditor evaluates whether controls are suitably designed (Type 1) and operating effectively (Type 2).
SRE and AI-agent fit notes
AI features in SOC 2 scope
AI features part of the service org's system are in SOC 2 scope:
- CC6.7 data transmission: encryption of prompts, outputs, training data.
- CC6.1-CC6.6 logical access: control over AI feature access.
- CC7.2 anomaly monitoring: agent behavior monitoring.
- CC7.3 incident response: AI-incident response procedures.
- CC8.1 change management: system prompt / tool scope versioning.
- CC9.1-CC9.2 vendor risk: model provider due diligence.
Privacy TSC for AI features processing personal data
When AI features process personal data, Privacy TSC adds:
- P3.0 consent (where applicable)
- P5.0 retention (training data, AI logs)
- P7.0 disclosure to third parties (model providers)
- P8.0 security
GDPR + SOC 2 Privacy TSC overlap substantial for orgs subject to both.
Stefan-context implementation sketch
- For US-customer-facing AI / SaaS engagements: SOC 2 TSC categories per customer demand.
- Map AI feature controls to TSC criteria.
- For combined SOC 2 + ISO 27001 implementations: single control set with dual mapping.