What ΔOS Is Not
Category boundaries and positioning clarity
Last reviewed: 2025-01-25
What ΔOS Is Not
Understanding what ΔOS is not is as important as understanding what it is. These boundaries are precise and definitional.
Misclassification leads to misconfigured expectations. This page prevents that.
Not an AI Safety Platform
ΔOS does not detect, prevent, or mitigate AI model safety issues. We govern actions, not model outputs.
| Concern | AI Safety Platform | ΔOS |
|---|---|---|
| Harmful content generation | Prevents | Not addressed |
| Model hallucination | Detects | Not addressed |
| Prompt injection | Blocks | Not addressed |
| Agent taking unauthorized action | Not addressed | Prevents |
What we do: Evaluate whether an agent should be allowed to execute a specific action, regardless of what the agent said or why.
Not an Observability Platform
ΔOS is not a logging, monitoring, or APM solution.
| Capability | Observability Platform | ΔOS |
|---|---|---|
| Log aggregation | Yes | No |
| Metrics collection | Yes | No |
| Distributed tracing | Yes | No |
| Alert on anomalies | Yes | No |
| Pre-execution policy gate | No | Yes |
What we do: Produce audit records as a byproduct of governance. These records are for accountability, not operational visibility.
Not a Security Tool
ΔOS does not perform threat detection, vulnerability scanning, or incident response.
| Function | Security Tool | ΔOS |
|---|---|---|
| Detect malware | Yes | No |
| Scan for vulnerabilities | Yes | No |
| SIEM correlation | Yes | No |
| Identity management | Yes | No |
| Policy-based action control | No | Yes |
What we do: Apply policy to agent-initiated actions. Security policies can be expressed in our framework, but we don't provide security intelligence.
Not a Workflow Orchestrator
ΔOS does not orchestrate, schedule, or manage workflows.
- Not Temporal: We don't manage state machines
- Not Airflow: We don't schedule DAGs
- Not Step Functions: We don't coordinate microservices
What we do: Sit at the judgment point of individual actions. We're a gate, not a choreographer.
Not an AI Agent Framework
ΔOS does not build, train, or run AI agents.
We integrate with agent frameworks (LangChain, AutoGPT, custom implementations) to provide governance. We're infrastructure they use, not a replacement for them.
Not an Approval Workflow Tool
ΔOS handles escalation to humans, but we're not an approval workflow product.
| Feature | Approval Workflow | ΔOS Escalation |
|---|---|---|
| Request routing | Complex rules | Policy-defined |
| Multi-level approval chains | Yes | No |
| SLA management | Yes | No |
| Substitute assignees | Yes | No |
| Real-time policy evaluation | No | Yes |
What we do: When policy requires human judgment, we surface that to the right human with full context. We don't manage approval hierarchies.
The Category We Define
Autonomy Governance Infrastructure: Deterministic policy evaluation for agent-initiated actions, with auditable judgment records and observable value attribution.
This category didn't exist before ΔOS. We're not competing with security tools, observability platforms, or workflow products. We're creating the infrastructure layer for governing autonomous agents.
Positioning Summary
| If you need to... | Don't use ΔOS | Do use |
|---|---|---|
| Filter harmful AI content | Yes | Guardrails, content filters |
| Monitor application performance | Yes | Datadog, New Relic |
| Detect security threats | Yes | CrowdStrike, Splunk |
| Orchestrate workflows | Yes | Temporal, Airflow |
| Build AI agents | Yes | LangChain, AutoGPT |
| Govern agent actions before execution | No | ΔOS |
See Also
- What is ΔOS? — What we actually are
- Autonomy Governance — The paradigm we define
- Intent — The unit of governance