AI risk is not measurable.
That’s the governance gap.

The Athena Noesis Decimal System (ANDS) provides a five-digit operational code for every AI system. It enables compliance, audit, and insurance teams to classify, govern, and verify AI risk using evidence—not marketing.

Why ANDS is necessary

Many AI systems now execute code, send messages, or access internal data—yet are described publicly as “assistants.” Without a consistent classification standard, organizations cannot distinguish low-risk tools from those with operational exposure.

  • Security and compliance teams cannot evaluate deployment risk consistently.
  • Insurance providers cannot price exposure accurately.
  • Regulators and auditors cannot confirm whether governance measures are adequate.

ANDS establishes a common operational language for describing and comparing AI systems, aligning technical reality with governance expectations.

The ANDS Code

Every system receives a five-number code representing its operational properties across five dimensions:

AxisDescription
CConnectivity — network exposure and integration
AAgency — level of autonomous or independent action
MMemory — persistence of stored data across sessions
GGovernance — oversight, audit, and control mechanisms
RRisk — potential for real-world consequence or harm

Example: 3.2.3.2.4 = moderately connected, bounded agency, persistent memory, basic governance, and material risk.

ANDS Classification Examples

The following table illustrates how the ANDS framework can be applied to known AI systems as of 2026. These ratings are based on publicly documented capabilities and governance structures, not private operational data.

System
click name for details
ANDS Code (C.A.M.G.R) Notes
ChatGPT (OpenAI)
ChatGPT (OpenAI) — General-Purpose Conversational AI

ChatGPT is a general conversational LLM used across many domains. Not inherently agentic — it responds to prompts rather than autonomously acting.

Key Observations for ANDS → ChatGPT
Connectivity: Moderate (internet access depends on deployment)
Agency: Low (no proactive autonomous operations by default)
Memory: Varies (session memory, sometimes persistent depending on plan/settings)
Governance: Moderate (centralized platform governance)
Risk: Moderate (wide capabilities but not autonomous)

Suggested ANDS Code: 3.2.2.3.3
3.2.2.3.3 Moderate connectivity and agency; persistent memory features under limited conditions; centralized governance and balanced operational risk.
Claude (Anthropic)
Claude (Anthropic) — Safety-Oriented Conversational AI

Claude focuses on safety and reasoning, maintaining discourse state longer and supporting deep tasks.

Key Observations for ANDS → Claude
Connectivity: Moderate (model hosted, limited external actions)
Agency: Low (no proactive autonomous behavior)
Memory: Moderate–High (persistent memory features introduced)
Governance: Strong (safety-focused design)
Risk: Lower to Moderate (safe design but still broad capabilities)

Suggested ANDS Code: 3.2.3.4.3
3.2.3.4.3 Low connectivity and agency; context-limited memory; strong governance model emphasizing constitutional constraints.
Gemini (Google DeepMind)
Gemini (Google DeepMind) — Multimodal Generative Model

Gemini is a multimodal LLM family (text, image, more) with real-time reasoning, large context windows, and integration into the Google ecosystem.

Key Observations for ANDS → Gemini
Connectivity: High (cloud, multimodal, search integration)
Agency: Low to Moderate (not inherently autonomous without orchestration)
Memory: Moderate (context windows can be large)
Governance: Moderate (platform controls and policies)
Risk: Elevated (wide use and capabilities)

Suggested ANDS Code: 3.3.3.3.4
3.3.3.3.4 High integration with Google ecosystem; cross-modal reasoning; moderate governance visibility; material risk in connected deployments.
Jules (Google)
Jules (Google) — Autonomous Coding Agent

Jules is an asynchronous agentic coding tool that reads your codebase, plans multi-step changes, and executes them securely, producing pull requests.

Key Observations for ANDS → Jules
Connectivity: High (needs internet/GitHub/Cloud VM)
Agency: High (executes tasks, not just suggestions)
Memory: Moderate (context of project state across runs)
Governance: Medium (human review/approval gates)
Risk: Elevated (real code changes in production-like environments)

Suggested ANDS Code: 4.4.3.3.4
4.4.3.3.4 Autonomous coding agent that performs asynchronous code tasks on GitHub; executes real actions with human review and governance oversight.

We also maintain a reference implementation with 1.3.3.5.2 classification. Contact us for details.

These classifications are subject to change as systems evolve or disclose additional operational evidence. The codes reflect relative exposure and governance assurance within the ANDS model, not performance or capability ratings.

ANDS Repository & Open Access Tools

The Athena Noesis Decimal System (ANDS) is provided as an open reference standard. Organizations, compliance teams, and insurers are invited to review, use, and extend the system freely.

Access the full source code, documentation, and reference implementation here:

Visit the ANDS Repository

Download the current reference script.

Contact

For implementation support or participation in verification pilots:

info@athena-noesis.com