AI Governance
Behavior Controls & Dual-Signature Overrides
Restrict the platform to models with verified training-opt-out and zero data retention. Any compliance-flagged snippet requires co-signature from a designated Compliance Officer — a developer alone cannot override the block.
Allowed Models
4/6
Opt-Out Verified
4
Leakage Risk
1
Pending Overrides
3
gpt-5.4 (Enterprise)
OpenAI
Training opt-out verified EU-WestZero retention ZDR-2026-0114
Allowed
Claude Opus 4.5 (Enterprise)
Anthropic
Training opt-out verified US-East30d retention ANT-ENT-9912
Allowed
Gemini 3.1 Pro (Vertex)
Google
Training opt-out verified EU-WestZero retention GCP-VTX-44A
Allowed
gpt-5.4 (Consumer API)
OpenAI
Opt-out denied — trains on inputs US (multi)30d retention —
Blocked by policy
Disabled
Llama 3.3 70B (self-hosted)
Meta
Training opt-out verified On-premZero retention INTERNAL-VPC-01
Allowed
Grok-3
xAI
Attestation pending US14d retention PENDING
Blocked by policy
Disabled