ChatGPT

TAS Score: S0/3 — D1/5 — A0 / T0 (S0: Pause ❌ — service controlled by OpenAI, can be discontinued or modified at any time. Exit ❌ — no way to export the model or fully replicate the service. Recoverability ❌ — no rollback, no local backup of model capabilities.) (D1: Personalisation ⚠️ builds user profiles. Urgency ⚠️ subscription pressure. Hidden cost ⚠️ data used for training. Transparency fragility ⚠️ value depends on proprietary model.)

Brief Description

Cloud-based AI assistant by OpenAI. GPT-4o, o1, and other proprietary models accessible via web interface, mobile apps, and API. The most widely used AI chatbot — and the benchmark against which self-hosted alternatives are measured.

Why it’s in the catalog

ChatGPT is the default AI tool for most people. Including it makes the autonomy trade-off visible: every feature ChatGPT offers has a self-hosted equivalent, but at a different point on the autonomy spectrum. This card exists for comparison, not recommendation.

Architectural Role

Cloud-hosted compute/inference. All processing happens on OpenAI’s infrastructure. No local component. API requires authentication and payment. Web interface requires account.

Technical Autonomy

  • Works without internet — no, cloud-only
  • Stores data locally — no, all data on OpenAI servers
  • Does not require external accounts — requires OpenAI account
  • Allows data export — conversation export exists, model cannot be exported
  • Provides offline updates — no

Philosophical Assessment (whose.world criteria)

Criterion Status Comments
Pause ❌ Cannot pause the service. OpenAI can modify, rate-limit, or discontinue at any time.
Exit ❌ Conversation history exportable. Model, capabilities, and fine-tuning cannot be taken.
Recoverability ❌ No rollback. Model versions change without user control.
Visibility ❌ Proprietary. No source code. No model weights. Architecture not disclosed.
External Dependencies ❌ Entirely dependent on OpenAI infrastructure and business decisions.

Configuration (Minimal) 

# API usage (requires API key and payment)
curl https://api.openai.com/v1/chat/completions \
  -H "Authorization: Bearer $OPENAI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"gpt-4o","messages":[{"role":"user","content":"Hello"}]}'

No self-hosted option exists.

Autonomous Alternatives

Alternative Autonomy Notes
Ollama + Open WebUI A3 / T2 Self-hosted ChatGPT-like experience. Free, private, offline.
llama.cpp A3 / T2 Low-level inference. Maximum control.
MiroThinker A3 / T2 Open-weight reasoning model for research tasks.

Trajectory

Direction: closing

OpenAI started as a non-profit research lab committed to open AI. GPT-2 was released with open weights. Since then: GPT-3, GPT-4, o1, o3 — all proprietary. Corporate restructuring from non-profit to capped-profit. Increasing prices, feature gating behind Plus/Pro/Team/Enterprise tiers.

Signal assessment:

Signal Status Evidence
License ⚠️ Proprietary. Moved from open (GPT-2) to fully closed.
Feature gating ⚠️ Free tier increasingly limited. Best models behind $20+/month subscription.
Self-hosting ⚠️ No self-hosted option. API-only access with rate limits and content policies.
Governance ⚠️ Corporate-controlled. Board drama. No community governance.

Signal key: ✅ opening · ➖ neutral · ⚠️ closing

Sources

This site uses Just the Docs, a documentation theme for Jekyll.