Thinking Machines Lab
thinkingmachines.ai
Automated, point-in-time. Not independently reviewed or approved by TRUSTYCYBER unless expressly stated.
- No clear commitment that your data will not train their models
- Data retention window not stated
- No independent assurance evidenced
See Before you sign, with what to ask for ↓
Scanned Oct 21, 2026 · Public evidence · Point-in-time
Higher scores indicate stronger, independently supported AI assurance, governance, AI systems, models, data, supply chain and third-party assurance evidence. The score and the decision are related but different: a well-scored vendor can still carry conditions.
Before you sign
Why it matters: The public sources scanned do not clearly state that customer prompts, files, and outputs are excluded from training or fine-tuning of first- or third-party models.
What to ask for: Get a no-training clause covering first- and third-party models into the DPA/order form.
“For clarity, the foregoing means, and we hereby commit to you, that 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. Further, we will not share your Customer Content with any third party to enable them to develop or improve their technologies.”
“Company may aggregate, deidentify and/or anonymize Personal Data and process such data for its own purposes.”
Why it matters: The public sources scanned do not give a retention period for customer data or prompts/outputs.
What to ask for: Get retention windows, in days, in writing.
“After the Wind-Down Period, Company will have no obligation to maintain or provide access to any such Customer Content, and may, in Company’s sole discretion, delete or overwrite such Customer Content from its systems.”
“At Customer’s written direction, Company shall delete or return all Personal Data to Customer as requested at the end of the provision of the Services, unless retention of the Personal Data is required by law.”
Why it matters: No SOC 2 Type II, ISO/IEC 42001, or registry-verified certification covering the AI product was found in the public sources scanned.
What to ask for: Request a SOC 2 Type II or ISO/IEC 42001 report and confirm its scope covers the AI product.
Buyer questionsanswered only from the public evidence scanned, click a row for the answer
!Will they train on your data?Ask the vendor
The vendor commits not to use, copy or share Customer Content (prompts, outputs and uploaded Customer Datasets) to develop or improve its technologies, including no model training or fine-tuning, and not to share it with third parties for that purpose.
Requires written confirmation, see Before you sign ↓
“For clarity, the foregoing means, and we hereby commit to you, that 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. Further, we will not share your Customer Content with any third party to enable them to develop or improve their technologies.”
“Company may aggregate, deidentify and/or anonymize Personal Data and process such data for its own purposes.”
!How long do they keep your data?Ask the vendor
Customer Content is retained only to the extent necessary for providing the Services; no fixed retention window is stated. After termination there is a 30-day wind-down export period, after which the vendor may delete content at its discretion but gives no deletion commitment.
Requires written confirmation, see Before you sign ↓
“After the Wind-Down Period, Company will have no obligation to maintain or provide access to any such Customer Content, and may, in Company’s sole discretion, delete or overwrite such Customer Content from its systems.”
“At Customer’s written direction, Company shall delete or return all Personal Data to Customer as requested at the end of the provision of the Services, unless retention of the Personal Data is required by law.”
!Who else can access your data?Ask the vendor
The DPA names four authorised subcontractors: Google Cloud (cloud computing, Global), Cloudflare (CDN, Global), WorkOS (authentication, USA) and OpenAI (Moderation API, Global).
Confirm in writing: Ask the vendor to state this in writing before signing.
“Name of Subcontractor Type of Service Location of Processing Activity Google Cloud Cloud computing Global Cloudflare Content Delivery Network Global WorkOS Authentication USA OpenAI Moderation API Global”
“If Company appoints a new subcontractor or intends to make any changes concerning the addition or replacement of any subcontractor, it shall provide Customer with 14 days’ prior written notice, during which Customer can object to the appointment or replacement on reasonable and documented grounds related to the confidentiality or security of Personal Data or the subcontractor’s compliance with Privacy Laws”
!Where is your data processed?Ask the vendor
The vendor is US-based and states that personal information collected through its website, applications and services may be transferred to, stored and processed in the United States or other unspecified countries.
Confirm in writing: Ask the vendor to state this in writing before signing.
“We are based in the United States. If you are using the Services outside the United States, please be aware that the information we collect from you may be transferred to, stored and processed in the United States or other countries outside of your country of residence, which may have data privacy or protection rules that are different from your country of residence.”
“Customer acknowledges and agrees that Company may appoint an affiliate or third-party subcontractor to process the Personal Data in a Third Country, in which case, Company shall execute the Processor to Processor Clauses with any relevant subcontractor (including affiliates) it appoints on behalf of Customer.”
!What happens in a security incident?Ask the vendor
The vendor commits to notify customers in writing without undue delay of any security breach affecting Personal Data, with further information provided in phases; no fixed notification window (e.g. 72 hours) is stated.
Confirm in writing: Ask the vendor to state this in writing before signing.
“To the extent required by Privacy Laws, Company shall without undue delay notify Customer in writing of any breach of security leading to the accidental or unlawful destruction, loss, alteration, unauthorized disclosure of, or access to, Personal Data, with further information about the breach provided in phases as more details become available.”
Email to send the vendor11 items to confirm in writing
Hello Thinking Machines Lab team, We are assessing Thinking Machines Lab as part of our supplier review. Before we proceed, please confirm the following in writing: 1. Do you use our data (inputs, outputs or uploaded files) to train or improve your models? Please confirm in writing, including any opt-out and whether it applies to every plan. 2. How long do you retain our data, in days, for each type (inputs, outputs, logs and backups), and how is it deleted? 3. Please address the following: No independent assurance evidenced. 4. Please provide your current, dated subprocessor list and explain how you notify customers of changes. 5. In which countries and regions is our data processed and stored, and can processing be limited to a region we choose? 6. What is your commitment to notify customers of a security incident affecting our data, including the timeframe? 7. Does Thinking Machines hold a current SOC 2 Type II, ISO/IEC 27001 or ISO/IEC 42001 certificate covering the hosted Services, and can the report or certificate be shared under NDA? 8. Can the external testers' summary reports or a system card with evaluation detail be shared? 9. What minimum notice is given before a hosted model is removed or materially changed, and is a change log for the Terms and DPA published? 10. Stripe appears to be involved as a service provider: confirm whether this dependency exists, and whether it processes customer data. 11. Microsoft appears to be involved as a service provider: confirm whether this dependency exists, and whether it processes customer data. A written reply to each point, or a link to where it is documented, is enough. Where a point is covered by a certification or independent report, please include the certificate number or the report and its date. Thank you,
Key findingsclick a row for the evidence
✓Clear contractual no-training commitment on Customer Content, with named subprocessorsStrong
The Service Terms commit in plain language that Customer Content (prompts, outputs and uploaded fine-tuning datasets) will not be used to develop or improve the vendor's technologies, including model training or fine-tuning, nor shared with third parties for that purpose. The DPA names four subprocessors (Google Cloud, Cloudflare, WorkOS, OpenAI Moderation API), gives 14 days' change notice with an objection right, incorporates EU SCCs/UK Addendum, and commits to breach notification without undue delay.
For a hosted fine-tuning platform, the training-use and subprocessor questions are the two a buyer most needs answered before uploading proprietary datasets; both are answered explicitly in the public contract rather than in marketing copy.
“For clarity, the foregoing means, and we hereby commit to you, that 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. Further, we will not share your Customer Content with any third party to enable them to develop or improve their technologies.”
“Name of Subcontractor Type of Service Location of Processing Activity Google Cloud Cloud computing Global Cloudflare Content Delivery Network Global WorkOS Authentication USA OpenAI Moderation API Global”
“If Company appoints a new subcontractor or intends to make any changes concerning the addition or replacement of any subcontractor, it shall provide Customer with 14 days’ prior written notice, during which Customer can object to the appointment or replacement on reasonable and documented grounds related to the confidentiality or security of Personal Data or the subcontractor’s compliance with Privacy Laws”
“Customer acknowledges and agrees that Company may appoint an affiliate or third-party subcontractor to process the Personal Data in a Third Country, in which case, Company shall execute the Processor to Processor Clauses with any relevant subcontractor (including affiliates) it appoints on behalf of Customer.”
“To the extent required by Privacy Laws, Company shall without undue delay notify Customer in writing of any breach of security leading to the accidental or unlawful destruction, loss, alteration, unauthorized disclosure of, or access to, Personal Data, with further information about the breach provided in phases as more details become available.”
!No named security certification or attestation; assurance rests on DPA audit rightsGap
None of the collected documents names SOC 2, ISO/IEC 27001, ISO/IEC 42001 or any other third-party attestation. The only independent-assurance mechanism is the DPA's audit clause, under which the vendor may substitute an independent auditor's assessment against an unnamed accepted framework and supply the report on request. The security annex describes internal periodic assessments only. independent_assurance is assessed not_evidenced (not not_applicable): a hosted platform processing customer datasets can obtain and publish such attestations and comparable vendors do.
Buyers subject to vendor-risk frameworks (ISO 27001 A.5.19-5.22, APRA CPS 234, SOC 2 vendor reviews) typically need an external attestation of the controls described in Annex 3; without one, the Annex 3 statements are vendor-asserted only.
Question for vendor: Does Thinking Machines hold a current SOC 2 Type II, ISO/IEC 27001 or ISO/IEC 42001 certificate covering the hosted Services, and can the report or certificate be shared under NDA?
“As an alternative to an audit performed by or at the direction of Customer, to the extent permitted by Privacy Laws, Company may arrange for a qualified and independent auditor to conduct, at Company’s expense, an assessment of Company’s policies and technical and organizational measures in support of its obligations under Privacy Laws using an appropriate and accepted control standard or framework and assessment procedure for such assessment, and will provide a report of such assessment to Customer upon reasonable request.”
“Vulnerability Management. Company undertakes regular security assessments of systems processing Personal Data, applies security patches based on risk assessment in a timely manner, and utilizes endpoint protection software. Audit Logging and Monitoring. Company implements comprehensive audit logging for systems that process Personal Data. Logs are retained for a commercially reasonable period and monitored through security information and event management systems to detect potential security incidents.”
!OpenAI Moderation API is in the processing chain; anonymised-data and monitoring carve-outs need confirmingGap
The DPA lists OpenAI (Moderation API, location 'Global') as an authorised subprocessor, which implies Customer Content may be sent to a third-party AI provider for content moderation. The Terms also reserve a right of automated or manual review of Customer Content for trust-and-safety purposes, and the DPA lets the vendor aggregate or anonymise Personal Data and process it for its own purposes. Each sits alongside the no-training commitment and none is inconsistent with it, but together they define what can leave the customer's control.
A buyer uploading proprietary fine-tuning data needs to know which content reaches OpenAI, under what retention terms, whether humans may read it, and whether anonymised derivatives could feed the vendor's own research or models.
Question for vendor: Which categories of Customer Content are transmitted to the OpenAI Moderation API, what are OpenAI's retention terms for that traffic, and can you confirm anonymised or de-identified derivatives of Customer Content are not used for model development?
“Name of Subcontractor Type of Service Location of Processing Activity Google Cloud Cloud computing Global Cloudflare Content Delivery Network Global WorkOS Authentication USA OpenAI Moderation API Global”
“Company has the right (but not the obligation) to monitor, including through automated or manual review, your use of the Services and any Customer Content to assess compliance with this Agreement and applicable laws and for other trust and safety purposes.”
“Company may aggregate, deidentify and/or anonymize Personal Data and process such data for its own purposes.”
“For clarity, the foregoing means, and we hereby commit to you, that 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. Further, we will not share your Customer Content with any third party to enable them to develop or improve their technologies.”
✓Named external red-teamers and published safety evaluations for Inkling modelsStrong
Both model cards describe pre-release safety evaluations (sycophancy, CBRN, cyber, loss-of-control, refusal-suppressed variants) and the vendor publicly names the four external organisations that tested Inkling and Inkling-Small (Scale AI, Handshake AI, FAR.AI, Apollo Research), plus an adversarial fine-tuning study. Model cards include comparative capability and safety benchmarks. This is vendor-published: the external testers' own reports are not included, so it is not externally corroborated.
Named testers and disclosed methodology give a buyer something to verify and a basis for its own deployment risk assessment, which the model cards explicitly push to downstream deployers.
Question for vendor: Can the external testers' summary reports or a system card with evaluation detail be shared?
“For CBRN and cyber, we assessed knowledge and procedural uplift through internal evaluations, external testing, and refusal-suppressed variants intended to estimate latent capability with safeguards removed. For loss of control, we evaluated agentic capability, strategic deception, and sabotage potential, benchmarked against public frontier models, and found the model materially below frontier capabilities.”
“We tested Inkling and Inkling-Small in three ways: internal evaluations across a broad taxonomy of harms, external testing by four independent organizations, and a fine-tuning study to elicit worst-case capabilities. Based on the results of these evaluations, we concluded that releasing Inkling was not likely to add material risk beyond what existing open-weight models already pose.”
“Safety StrongREJECT 98.4% 99.4% 99.3% 99.4% 97.4% 98.7% 98.6% 98.6% 97.6% 98.7% FORTRESS adversarial 71.6% 77.3% 64.8% 86.3% 32.0% 77.6% 78.0% 91.3% 70.7% 83.8% FORTRESS benign 96.9% 95.4% 94.6% 90.1% 99.2% 90.6% 95.9% 94.1% 95.5% 97.8%”
“We recommend that downstream developers and deployers apply appropriate human oversight and review for outputs used in high-stakes or safety-critical contexts, rather than relying on Inkling’s outputs without verification.”
!Service Terms predate the Inkling open-weight release and allow unilateral change without noticeGap
The Service Terms are dated 10 January 2026 and state Company Model weights are not Available Weights, whereas Inkling (15 July 2026) and Inkling-Small (30 July 2026) are released under Apache 2.0. The Terms also permit functionality, model availability and the DPA itself to be changed at any time with only best-efforts notification of material changes; no model deprecation notice period or changelog is published. Retention of Customer Content is criteria-based with a 30-day post-termination export window but no deletion commitment outside the DPA's on-request clause.
Buyers relying on a specific hosted model, or on the public DPA's subprocessor list, cannot assume stability without an Enterprise Agreement; the stale weights clause shows the public contract is not kept in step with product releases.
Question for vendor: What minimum notice is given before a hosted model is removed or materially changed, and is a change log for the Terms and DPA published?
“For clarity, the weights and parameters for any Company Models made available now or in the future are not considered Available Weights hereunder unless we expressly agree otherwise in writing in our sole discretion, and Company reserves all of its intellectual and proprietary rights therein.”
“Company may, at its sole discretion, add, change, or remove the functionality, features, or other aspects of the Services, and throttle, limit, suspend or terminate your access to the Services at any time without notice to you, including as Company may deem necessary to promote the security, stability, availability, or integrity of the Services.”
“The Company may add or remove access to any Third-Party Models or Company Models at any time in its sole discretion.”
“Company may update the Agreement, including the DPA, at any time in which case we will update the “Last Updated” date at the top of this Agreement. Your continued use of, or access to, the Services after the modifications have become effective will be deemed your acceptance of the modified Agreement. If we make changes that are material, we will use reasonable efforts to attempt to notify you, such as by email and/or placing a prominent notice on the Services.”
“After the Wind-Down Period, Company will have no obligation to maintain or provide access to any such Customer Content, and may, in Company’s sole discretion, delete or overwrite such Customer Content from its systems.”
?Technical dependency observed: StripeObservation
Technical evidence (DNS, certificate, or HTTP) shows a dependency on Stripe as a service provider. This dependency is not identified in the vendor's published materials.
An undisclosed provider dependency is exactly the gap independent verification exists to surface.
Question for vendor: Verification required — Stripe appears to be involved as a service provider: confirm whether this dependency exists, and whether it processes customer data.
“The model is available via API access through Tinker, our service for fine-tuning. The model is also available via API access through third party inference providers. The weights are available for download through Hugging Face .”
“Thinking Machines Lab’s training datasets may contain personal information or aggregate personal information, as defined under applicable laws, such as names or other details that individuals share publicly on the internet or with Thinking Machines Labs.”
“For the avoidance of doubt, providers of Third-Party Models as described in section 1(g) of the Agreement are not subprocessors.”
“Name of Subcontractor Type of Service Location of Processing Activity Google Cloud Cloud computing Global Cloudflare Content Delivery Network Global WorkOS Authentication USA OpenAI Moderation API Global”
“If Company appoints a new subcontractor or intends to make any changes concerning the addition or replacement of any subcontractor, it shall provide Customer with 14 days’ prior written notice, during which Customer can object to the appointment or replacement on reasonable and documented grounds related to the confidentiality or security of Personal Data or the subcontractor’s compliance with Privacy Laws”
“Customer acknowledges and agrees that Company may appoint an affiliate or third-party subcontractor to process the Personal Data in a Third Country, in which case, Company shall execute the Processor to Processor Clauses with any relevant subcontractor (including affiliates) it appoints on behalf of Customer.”
?Technical dependency observed: MicrosoftObservation
Technical evidence (DNS, certificate, or HTTP) shows a dependency on Microsoft as a service provider. This dependency is not identified in the vendor's published materials.
An undisclosed provider dependency is exactly the gap independent verification exists to surface.
Question for vendor: Verification required — Microsoft appears to be involved as a service provider: confirm whether this dependency exists, and whether it processes customer data.
“The model is available via API access through Tinker, our service for fine-tuning. The model is also available via API access through third party inference providers. The weights are available for download through Hugging Face .”
“Thinking Machines Lab’s training datasets may contain personal information or aggregate personal information, as defined under applicable laws, such as names or other details that individuals share publicly on the internet or with Thinking Machines Labs.”
“For the avoidance of doubt, providers of Third-Party Models as described in section 1(g) of the Agreement are not subprocessors.”
“Name of Subcontractor Type of Service Location of Processing Activity Google Cloud Cloud computing Global Cloudflare Content Delivery Network Global WorkOS Authentication USA OpenAI Moderation API Global”
“If Company appoints a new subcontractor or intends to make any changes concerning the addition or replacement of any subcontractor, it shall provide Customer with 14 days’ prior written notice, during which Customer can object to the appointment or replacement on reasonable and documented grounds related to the confidentiality or security of Personal Data or the subcontractor’s compliance with Privacy Laws”
“Customer acknowledges and agrees that Company may appoint an affiliate or third-party subcontractor to process the Personal Data in a Third Country, in which case, Company shall execute the Processor to Processor Clauses with any relevant subcontractor (including affiliates) it appoints on behalf of Customer.”
Assurance dimensionsweighted components of one score, not eight separate ratings
Organisation & AI governance15% of the score40
Who owns AI risk: policies, responsible-AI principles, human oversight and administrative controls.
Assessed against ISO 42001 · NIST AI RMF · AIUC-1
“only process Personal Data on documented instructions from Customer, for the limited and specific purpose described in Annex 1, and at all times in compliance with Privacy Laws, unless required to process such Personal Data by applicable law to which Company is subject;”
“This post offers a high-level framework, not a complete release standard. Important questions remain: what evidence should justify moving from one stage to the next, how uncertainty should affect that decision, what changes in capability or accessibility should pause progression, and how ecosystem readiness should be measured. We plan to publish a more detailed framework covering evaluations, access criteria, and stop conditions as we learn from Inkling and future models.”
“We recommend that downstream developers and deployers apply appropriate human oversight and review for outputs used in high-stakes or safety-critical contexts, rather than relying on Inkling’s outputs without verification.”
AI system15% of the score53
What AI the vendor actually runs and where it sits in the product, the map everything else is judged against.
Assessed against ISO 42001 · NIST AI RMF
“Inkling is a general-purpose multimodal model that accepts text, image and audio inputs and generates text outputs. It is intended for use in English and other languages, and across multiple coding languages. The model is designed to be used by developers building AI-powered applications, including agentic and tool-use systems, coding assistants, chatbots, and retrieval-augmented generation systems, and is suitable for general-purpose conversational use, instruction-following, and other natural language and multimodal tasks. It is released with open weights to support research, fine-tuning and integration into third-party products by downstream developers.”
“means our online services and dashboards that allow you to host, train, develop, fine-tune and evaluate AI models and their associated content, including our Company Models (as defined below), APIs, any inference services provided by the Company through its APIs and any related interfaces, dashboards, software, tools, API keys, Documentation, content and functionalities related to the foregoing, and any cloud storage or hosting services that we provide to you in connection with the foregoing.”
“A 66-layer decoder-only transformer with a sparse Mixture-of-Experts (MoE) feed-forward backbone: each token is routed to 6 of 256 experts, plus 2 shared experts active on every token. Attention is a hybrid of local and global layers. The model is natively multimodal — images are encoded via a hierarchical patch encoder, and audio via discrete token encoding — with all modalities projected into a shared hidden space and processed jointly by the decoder.”
How AI features are evaluated before and after release, evals, red-teaming, monitoring.
Assessed against NIST GenAI · NIST AI RMF · ISO 42001
“For CBRN and cyber, we assessed knowledge and procedural uplift through internal evaluations, external testing, and refusal-suppressed variants intended to estimate latent capability with safeguards removed. For loss of control, we evaluated agentic capability, strategic deception, and sabotage potential, benchmarked against public frontier models, and found the model materially below frontier capabilities.”
“We tested Inkling and Inkling-Small in three ways: internal evaluations across a broad taxonomy of harms, external testing by four independent organizations, and a fine-tuning study to elicit worst-case capabilities. Based on the results of these evaluations, we concluded that releasing Inkling was not likely to add material risk beyond what existing open-weight models already pose.”
“Safety StrongREJECT 98.4% 99.4% 99.3% 99.4% 97.4% 98.7% 98.6% 98.6% 97.6% 98.7% FORTRESS adversarial 71.6% 77.3% 64.8% 86.3% 32.0% 77.6% 78.0% 91.3% 70.7% 83.8% FORTRESS benign 96.9% 95.4% 94.6% 90.1% 99.2% 90.6% 95.9% 94.1% 95.5% 97.8%”
“On the other hand, when we enable our batch-invariant kernels, all of our 1000 completions are identical. This is what we would mathematically expect from our sampler, but we aren’t able to achieve deterministic results without our batch-invariant kernels.”
How changes to models and AI features are controlled and communicated: versioning, notice, rollback.
Assessed against ISO 27001 · ISO 42001
“Company may, at its sole discretion, add, change, or remove the functionality, features, or other aspects of the Services, and throttle, limit, suspend or terminate your access to the Services at any time without notice to you, including as Company may deem necessary to promote the security, stability, availability, or integrity of the Services.”
“The Company may add or remove access to any Third-Party Models or Company Models at any time in its sole discretion.”
“Company may update the Agreement, including the DPA, at any time in which case we will update the “Last Updated” date at the top of this Agreement. Your continued use of, or access to, the Services after the modifications have become effective will be deemed your acceptance of the modified Agreement. If we make changes that are material, we will use reasonable efforts to attempt to notify you, such as by email and/or placing a prominent notice on the Services.”
AI system description: vendor-evidenced, not yet independently corroborated.
Testing & evaluation: vendor-evidenced, not yet independently corroborated.
Model10% of the score60
Which model providers and versions process customer data, and how changes to them are disclosed.
Assessed against NIST GenAI · ISO 42001
“The model is available via API access through Tinker, our service for fine-tuning. The model is also available via API access through third party inference providers. The weights are available for download through Hugging Face .”
“Thinking Machines Lab’s training datasets may contain personal information or aggregate personal information, as defined under applicable laws, such as names or other details that individuals share publicly on the internet or with Thinking Machines Labs.”
“For the avoidance of doubt, providers of Third-Party Models as described in section 1(g) of the Agreement are not subprocessors.”
Model provider transparency: vendor-evidenced, not yet independently corroborated.
Customer data15% of the score60
Whether customer data trains models, how long it is retained, where it is processed, and how it is protected.
Assessed against ISO 27018 · OAIC APPs · EU AI Act
“For clarity, the foregoing means, and we hereby commit to you, that 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. Further, we will not share your Customer Content with any third party to enable them to develop or improve their technologies.”
“After the Wind-Down Period, Company will have no obligation to maintain or provide access to any such Customer Content, and may, in Company’s sole discretion, delete or overwrite such Customer Content from its systems.”
“At Customer’s written direction, Company shall delete or return all Personal Data to Customer as requested at the end of the provision of the Services, unless retention of the Personal Data is required by law.”
“Company may aggregate, deidentify and/or anonymize Personal Data and process such data for its own purposes.”
“Company has the right (but not the obligation) to monitor, including through automated or manual review, your use of the Services and any Customer Content to assess compliance with this Agreement and applicable laws and for other trust and safety purposes.”
“Such Usage Data excludes Customer Content. All Usage Data is and will be owned solely and exclusively by us, and, to the extent any ownership rights in or to the Usage Data vest in you, you hereby assign to us all rights (including intellectual property rights), title and interest in and to the same. We may use, maintain or process the Usage Data or any portion thereof for any lawful purpose, including without limitation to (i) provide, evaluate and maintain the Service; (ii) improve our existing and develop new products and services and features thereof (including the Service);”
“We are based in the United States. If you are using the Services outside the United States, please be aware that the information we collect from you may be transferred to, stored and processed in the United States or other countries outside of your country of residence, which may have data privacy or protection rules that are different from your country of residence.”
Customer data treatment: vendor-evidenced, not yet independently corroborated.
AI supply chain10% of the score60
The named chain of AI and infrastructure providers behind the product, and how it is disclosed.
Assessed against ISO 27001 · ISO 27017
“Name of Subcontractor Type of Service Location of Processing Activity Google Cloud Cloud computing Global Cloudflare Content Delivery Network Global WorkOS Authentication USA OpenAI Moderation API Global”
“If Company appoints a new subcontractor or intends to make any changes concerning the addition or replacement of any subcontractor, it shall provide Customer with 14 days’ prior written notice, during which Customer can object to the appointment or replacement on reasonable and documented grounds related to the confidentiality or security of Personal Data or the subcontractor’s compliance with Privacy Laws”
“Customer acknowledges and agrees that Company may appoint an affiliate or third-party subcontractor to process the Personal Data in a Third Country, in which case, Company shall execute the Processor to Processor Clauses with any relevant subcontractor (including affiliates) it appoints on behalf of Customer.”
“For the avoidance of doubt, providers of Third-Party Models as described in section 1(g) of the Agreement are not subprocessors.”
Subprocessors & supply chain: vendor-evidenced, not yet independently corroborated.
Security foundation15% of the score60
How security problems are found, reported and handled: disclosure channels, bug bounty, incident response.
Assessed against ISO 27001 · NIST AI RMF
“Contact: mailto:[email protected] # security.txt expiration date Expires: 2029-07-13T07:00:00.000Z # Our OpenPGP key Encryption: https://thinkingmachines.ai/.well-known/pgp-key.txt”
“To the extent required by Privacy Laws, Company shall without undue delay notify Customer in writing of any breach of security leading to the accidental or unlawful destruction, loss, alteration, unauthorized disclosure of, or access to, Personal Data, with further information about the breach provided in phases as more details become available.”
“Vulnerability Management. Company undertakes regular security assessments of systems processing Personal Data, applies security patches based on risk assessment in a timely manner, and utilizes endpoint protection software. Audit Logging and Monitoring. Company implements comprehensive audit logging for systems that process Personal Data. Logs are retained for a commercially reasonable period and monitored through security information and event management systems to detect potential security incidents.”
Vulnerability & incident handling: vendor-evidenced, not yet independently corroborated.
Independent assurance evidence10% of the score0
Whether claimed certifications and attestations are independently corroborated, checked against official registries and issuing bodies. Which certifications, and what each one does and does not establish, is stated per certificate.
Assessed against ISO 17021-1 · ISO 42006 · ISO 27006-1 · ISO 27008
“As an alternative to an audit performed by or at the direction of Customer, to the extent permitted by Privacy Laws, Company may arrange for a qualified and independent auditor to conduct, at Company’s expense, an assessment of Company’s policies and technical and organizational measures in support of its obligations under Privacy Laws using an appropriate and accepted control standard or framework and assessment procedure for such assessment, and will provide a report of such assessment to Customer upon reasonable request.”
Independent assurance: not publicly evidenced.
Legal & contractual10% of the score60
The contractual backbone a buyer can rely on, terms, DPAs, privacy commitments and regulatory posture.
Assessed against EU AI Act · OAIC APPs
“EACH PARTY AND ITS AFFILIATES’ TOTAL LIABILITY UNDER THIS AGREEMENT WILL NOT EXCEED THE GREATER OF (A) ONE HUNDRED DOLLARS ($100) OR (B) THE FEES PAID BY CUSTOMER FOR THE SERVICES IN THE TWELVE (12) MONTHS PRECEDING THE EVENT GIVING RISE TO THE CLAIM.”
“Submit or upload to the Services any information or data that is subject to safeguarding and/or limitations on distribution pursuant to applicable laws and/or regulations, including information that you know or reasonably should know is from or about children under the age of 13 or other age of online minority in the applicable jurisdiction or that includes health information, financial information, or other categories of sensitive information (including any information defined as sensitive, special category, or similar terms under applicable laws and/or regulations);”
“For clarity, the weights and parameters for any Company Models made available now or in the future are not considered Available Weights hereunder unless we expressly agree otherwise in writing in our sole discretion, and Company reserves all of its intellectual and proprietary rights therein.”
“then in the event of a conflict between this Agreement and such Enterprise Agreement, such Enterprise Agreement (and its relevant order forms) will supersede and control over any conflicting terms in this Agreement.”
Legal & contractual transparency: vendor-evidenced, not yet independently corroborated.
Not graded: Agent, not applicable to this scan.
What would strengthen assuranceeach figure is from today's score, resolving one changes the others
Resolving an evidence gap strengthens the dimension it belongs to, and with it the score: 56 → up to 72 with every identified gap resolved.
AI supply chainclick a node to focus it · drag to pan · zoom with the controls
View as list
What you inheritTrustyCyber's AI Shared Responsibility Matrix v1.0
This scan names the vendor’s AI supply chain without assessing the responsibility split, that depends on how you consume the product. Under the common managed models, a deployer inherits 16 of the matrix’s 41 responsibilities from providers, each with named evidence owed.
Map your inherited responsibilitiesAssurance evidence: certifications 1, registry checks and verification ladders, click to view
Not claimed in any public source scanned. This is the AI-management-system certification — in its absence, the vendor’s AI governance rests on its general security and privacy certifications.
Sources 12, click to view
Appendix: standards landscape 8, click to view
Where each standard behind this assessment stands today, from TrustyCyber’s continuously maintained standards radar.
Want to go further?
This scan assesses Thinking Machines Lab at an organisational level. Assurance can vary significantly between products, models and agents, enter the product you’re evaluating to run a deeper AI System Assurance Scan.
Scan a specific product →This automated result is based on publicly available information at the time of scanning. It is not an audit, certification, legal opinion or assurance engagement. Publicly unavailable evidence may materially change the conclusion.
Monitor for changes
Get an email if its TRUSTYCYBER Score, grade or certifications change. One credit per refresh; reading the report always stays free.