Reviewed Jun 2026 · some cells re-verified Aug 2026 · policies are checked weekly for changes · stated — every cell is read from Thinking Machines's published policy; none has been tested (run by us) yet.
2Trains AIstated
The Service Terms state “we will not use, copy or share any of your Customer Content (including your Customer Datasets) to develop or improve our technologies. For instance, we will not use your Customer Content for model training or fine-tuning” — no-train by default, with no free or beta carve-out. The separate Free Research API tier (served via OpenRouter) does train on Prompts and Outputs; it is a different service under its own terms.
Thinking Machines Service Terms of Use — Customer Content · accessed Aug 2026 ↗
0API access — feeds Agent Readiness, not the /12stated
Tinker's API serves fine-tuning and inference; there is no consumer-grantable OAuth API for an agent to fetch your own account or Playground data — the self-data path is an email request to privacy@thinkingmachines.ai.
Tinker Documentation (developer-facing) · accessed Aug 2026 ↗
Graded 2026-08-31 against the Thinking Machines Website Privacy Notice (last updated May 19, 2026) and the Service Terms of Use (last updated Jan 10, 2026). Consumer surface = the Inkling Playground in the Tinker console, operated by Thinking Machines Lab, Inc., a US company. Every privacy right runs through privacy@thinkingmachines.ai — no portal, no export tool, no stated response timeline. The Service Terms carry an unqualified no-train commitment covering all Prompts and Outputs. A second lane exists and grades the other way: the TML Free Research API Tier (Inkling served free via OpenRouter, the only Exhibit B platform; ToS and Privacy Notice both last updated Aug 20, 2026) takes a worldwide, irrevocable, perpetual licence to Prompts and Outputs “to train, fine-tune, evaluate, and improve the Company Models and other AI models”, retained “during and after the Term”, with no stated opt-out. That endpoint is outside the graded surface; the Trains cell scores the Playground. This company also publishes its models as open weights. Running them yourself, or through another host, is not governed by this policy — this grade describes Thinking Machines' own service. All cells stated, not tested.
Events
Breaches, policy changes, and export-format changes we’ve recorded at Thinking Machines — other parties’ actions, alongside the enforcement record above. An event never changes a grade by itself. All events →
No events on record for Thinking Machines.
Grade history
Every change to this company’s grade since it was first published, and why. All grade changes →
No changes since this grade was first published in Jun 2026.