Cohere
cohere.com
Automated, point-in-time. Not independently reviewed or approved by TRUSTYCYBER unless expressly stated.
- No formal subprocessor register disclosed
- No clear commitment that your data will not train their models
- Data retention window not stated
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 vendor names AI-related providers or partially addresses its supply chain, but the public sources scanned do not include a formal, dated subprocessor register.
What to ask for: Request a dated subprocessor register with change-notification terms for the DPA/order form.
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.
“Cohere does not have access to data processed in connection with the use of Cohere models or products in Private Deployments or Deployments on Third-Party Managed AI services (e.g. on AWS, Azure, or GCP). The data remains within the enterprise customer’s or the third-party cloud provider’s controlled environment and is never accessible to or used by Cohere (for example, we do not train our models on it).”
“we collect content you submit to the Cohere Products, and outputs generated by the Cohere Products. We may use this input and output data to conduct research and development. This includes analyzing usage trends and using inputs/outputs to improve the performance and safety of Cohere Products.”
“However, as a further safeguard, Cohere takes steps to de-identify input and output data prior to any use for research and development purposes. If you need to include personal information in inputs to Cohere Products, consider moving to an enterprise tier and entering into a commercial agreement and DPA with Cohere.”
“Maintain full control over your data with customizable handling and retention settings. Opt out of model training at any time — your data stays yours.”
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.
“Enterprise Users of the Platform can consult retention practices for inputs and outputs on the Platform here . Retention of inputs and outputs on the Platform is generally 30 days for Enterprise Users.”
Buyer questionsanswered only from the public evidence scanned, click a row for the answer
!Will they train on your data?Ask the vendor
For Trial Users and Researchers on the hosted Platform, Cohere collects submitted inputs and generated outputs and may use them for research and development, including to improve the performance and safety of Cohere Products.
Requires written confirmation, see Before you sign ↓
“Cohere does not have access to data processed in connection with the use of Cohere models or products in Private Deployments or Deployments on Third-Party Managed AI services (e.g. on AWS, Azure, or GCP). The data remains within the enterprise customer’s or the third-party cloud provider’s controlled environment and is never accessible to or used by Cohere (for example, we do not train our models on it).”
“we collect content you submit to the Cohere Products, and outputs generated by the Cohere Products. We may use this input and output data to conduct research and development. This includes analyzing usage trends and using inputs/outputs to improve the performance and safety of Cohere Products.”
“However, as a further safeguard, Cohere takes steps to de-identify input and output data prior to any use for research and development purposes. If you need to include personal information in inputs to Cohere Products, consider moving to an enterprise tier and entering into a commercial agreement and DPA with Cohere.”
“Maintain full control over your data with customizable handling and retention settings. Opt out of model training at any time — your data stays yours.”
!How long do they keep your data?Ask the vendor
Retention of inputs and outputs on the hosted Platform is generally 30 days for Enterprise Users; a separate retention page is referenced for detail.
Requires written confirmation, see Before you sign ↓
“Enterprise Users of the Platform can consult retention practices for inputs and outputs on the Platform here . Retention of inputs and outputs on the Platform is generally 30 days for Enterprise Users.”
!Who else can access your data?Ask the vendor
Cohere may use its own AI or unnamed third-party AI tools to analyse the contents of support tickets, and asks users not to include sensitive, confidential or personal information in tickets.
Requires written confirmation, see Before you sign ↓
“We may use our own AI technology or third-party AI tools to analyze the contents of support tickets. This helps us summarize inquiries, draft responses, and resolve issues faster. Please do not include sensitive, confidential, or personal information in your support tickets.”
“Our cloud infrastructure provided by GCP in the United States, for instance, is protected under the EU-US Data Privacy Framework , including its UK extension.”
!Where is your data processed?Ask the vendor
Cohere transfers personal information to Canada, the United States and the United Kingdom among other countries, uses adequacy decisions or Standard Contractual Clauses where required, and discloses that foreign governments or courts may in some circumstances be entitled to access it.
Confirm in writing: Ask the vendor to state this in writing before signing.
“Examples of countries we transfer personal information to include, but are not limited to, Canada, the United States, and the United Kingdom. As a result, in certain circumstances, other foreign governments, courts, law enforcement agencies, or regulatory agencies may be entitled to access the personal information collected and held by Cohere, its affiliates, partners, and service providers.”
“Users of the Platform or Cohere APIs can consult our list of sub-processors , which includes data processing locations.”
!What happens in a security incident?Ask the vendor
Cohere operates a Bug Bounty program and publishes a Responsible Disclosure Policy for reporting potential vulnerabilities; continuous threat detection monitoring is also claimed for the API platform.
Confirm in writing: Ask the vendor to state this in writing before signing.
“The Cohere Bug Bounty program is designed to recognize the community’s support in protecting our users’ privacy and security. To learn more about the program, download our Responsible Disclosure Policy .”
“Our advanced threat detection systems continuously monitor for suspicious activity”
Email to send the vendor7 items to confirm in writing
Hello Cohere team, We are assessing Cohere as part of our supplier review. Before we proceed, please confirm the following in writing: 1. Please provide your current, dated subprocessor list and explain how you notify customers of changes. 2. 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. 3. How long do you retain our data, in days, for each type (inputs, outputs, logs and backups), and how is it deleted? 4. In which countries and regions is our data processed and stored, and can processing be limited to a region we choose? 5. What is your commitment to notify customers of a security incident affecting our data, including the timeframe? 6. Please share the current SOC 2 Type II report, and the ISO/IEC 27001 and ISO/IEC 42001 certificates with their scope statements, certification bodies and expiry dates. 7. 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
✓Private and third-party cloud deployments keep Cohere out of the data pathStrong
Cohere states plainly that for Private Deployments and deployments on AWS, Azure or GCP managed AI services it has no access to processed data and does not train on it. On its own hosted Platform, Enterprise User inputs and outputs are retained for generally 30 days and training can be opted out of, with Customer Data governed by the enterprise agreement and DPA.
A buyer who deploys privately or via a hyperscaler marketplace can rely on a clear no-access, no-training position; one using the hosted API has a short, stated retention window and contractual data commitments to point to.
“Cohere does not have access to data processed in connection with the use of Cohere models or products in Private Deployments or Deployments on Third-Party Managed AI services (e.g. on AWS, Azure, or GCP). The data remains within the enterprise customer’s or the third-party cloud provider’s controlled environment and is never accessible to or used by Cohere (for example, we do not train our models on it).”
“Enterprise Users of the Platform can consult retention practices for inputs and outputs on the Platform here . Retention of inputs and outputs on the Platform is generally 30 days for Enterprise Users.”
“Maintain full control over your data with customizable handling and retention settings. Opt out of model training at any time — your data stays yours.”
“Cohere’s use of Customer Data is governed by the terms and conditions of the enterprise agreement and, where applicable, the Data Processing Addendum entered into between the parties.”
!Hosted-tier training is opt-out, and trial/research inputs are used for R&DGap
The privacy policy discloses that Trial User and Researcher inputs and outputs may be used to improve Cohere Products (with de-identification as a safeguard), and the security page frames enterprise training as something customers 'opt out of at any time' rather than a default-off commitment. The Enterprise Data Commitments and Model Training Privacy Notice that would settle the enterprise default were not in the collected documents.
Whether customer prompts and outputs feed model improvement depends on tier and on an opt-out the buyer must actively exercise; evaluation on a free trial account carries explicit training use.
Question for vendor: For a paid enterprise account on the hosted Cohere API Platform, is training on Customer Data off by default, and can you provide the Enterprise Data Commitments and DPA language that states this?
“we collect content you submit to the Cohere Products, and outputs generated by the Cohere Products. We may use this input and output data to conduct research and development. This includes analyzing usage trends and using inputs/outputs to improve the performance and safety of Cohere Products.”
“However, as a further safeguard, Cohere takes steps to de-identify input and output data prior to any use for research and development purposes. If you need to include personal information in inputs to Cohere Products, consider moving to an enterprise tier and entering into a commercial agreement and DPA with Cohere.”
“Maintain full control over your data with customizable handling and retention settings. Opt out of model training at any time — your data stays yours.”
!SOC 2, ISO 27001 and ISO 42001 claimed without scope, issuer or datesGap
SOC 2 Type II is claimed for the API platform and SOC 2 Type 2, ISO 27001 and ISO 42001 for North, alongside annual third-party audits and penetration tests. None of the collected pages names the auditor or certification body, the certified scope, or the report or certificate dates; the Trust Center that likely holds them was referenced but not collected.
ISO 42001 in particular is a differentiating AI management system claim; without scope and issuer it cannot be relied on for a vendor assessment. All assurance here is vendor-asserted.
Question for vendor: Please share the current SOC 2 Type II report, and the ISO/IEC 27001 and ISO/IEC 42001 certificates with their scope statements, certification bodies and expiry dates.
“Our API platform is hosted on industry-leading cloud infrastructure and is SOC 2 Type II compliant”
“Exceeding global security standards: with SOC 2 Type 2, ISO 27001, and ISO 42001 certifications.”
“We conduct annual third-party audits and penetration tests to identify and address vulnerabilities”
!Unnamed third-party AI tools and an uncollected sub-processor listGap
Cohere discloses GCP in the United States as its cloud infrastructure and NVIDIA hardware for training and inference, and refers users to a sub-processor list with processing locations that was not in the collected set. It also says unnamed third-party AI tools may analyse support tickets, and North is model agnostic so a customer-supplied model may sit in the request path.
The request-path recipients for hosted use are only partly visible from the collected documents, and support-ticket content may reach an unnamed AI provider.
Question for vendor: Which third-party AI tools process support ticket contents, and can you provide the current sub-processor list with processing locations?
“We may use our own AI technology or third-party AI tools to analyze the contents of support tickets. This helps us summarize inquiries, draft responses, and resolve issues faster. Please do not include sensitive, confidential, or personal information in your support tickets.”
“Our cloud infrastructure provided by GCP in the United States, for instance, is protected under the EU-US Data Privacy Framework , including its UK extension.”
“Users of the Platform or Cohere APIs can consult our list of sub-processors , which includes data processing locations.”
“Bring your organization’s existing models: North is model agnostic, use our generative model offerings or bring your favorites.”
“Additionally, our collaboration with NVIDIA finds that North generates more insights, for fewer tokens, with Cohere models trained and deployed on NVIDIA Blackwell and Hopper.”
!Dated model retirement notices, but no customer incident-notification commitment foundGap
Release notes carry dated retirement notices with replacement guidance and a Deprecations page, and North 2 logs all changes — good change-management hygiene. However, the vulnerability domain is covered only by a bug bounty, responsible disclosure policy, threat monitoring and pen testing; no breach-notification commitment or timeline to customers appears in the collected documents (it may sit in the uncollected DPA or SLO).
A buyer needs to know how and when Cohere notifies customers of a security incident affecting their data; this is normally a DPA term and should be confirmed in writing.
Question for vendor: What is your contractual commitment and timeline for notifying customers of a security incident affecting Customer Data?
“Effective April 4, 2026, the following models are no longer available. Requests using these model IDs will fail.”
“System observability: Gain visibility into North's operations with easy-to-use tools for monitoring service performance and detailed logs of all changes.”
“The Cohere Bug Bounty program is designed to recognize the community’s support in protecting our users’ privacy and security. To learn more about the program, download our Responsible Disclosure Policy .”
“Our advanced threat detection systems continuously monitor for suspicious activity”
?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.
“Bring your organization’s existing models: North is model agnostic, use our generative model offerings or bring your favorites.”
“We may use our own AI technology or third-party AI tools to analyze the contents of support tickets. This helps us summarize inquiries, draft responses, and resolve issues faster. Please do not include sensitive, confidential, or personal information in your support tickets.”
“Our cloud infrastructure provided by GCP in the United States, for instance, is protected under the EU-US Data Privacy Framework , including its UK extension.”
“Users of the Platform or Cohere APIs can consult our list of sub-processors , which includes data processing locations.”
“Additionally, our collaboration with NVIDIA finds that North generates more insights, for fewer tokens, with Cohere models trained and deployed on NVIDIA Blackwell and Hopper.”
Assurance dimensionsweighted components of one score, not eight separate ratings
Organisation & AI governance15% of the score45
Who owns AI risk: policies, responsible-AI principles, human oversight and administrative controls.
Assessed against ISO 42001 · NIST AI RMF · AIUC-1
“AI Governance Safety Modes Usage Policy Model Technical Documentation Cohere’s AI Governance Approach Secure AI Frontier Model Framework”
“Defined roles and responsibilities for personnel with respect to the treatment of personal information. Processes to respond to data subject requests and complaints in a timely and effective manner.”
“Autonomy policies: Build agents designed to only take actions they are authorized to take, and seek human oversight for critical decisions or actions.”
“As we explain in that notice, as an enterprise AI company, Cohere does not intentionally collect personal information for training purposes.”
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
“North 2 is Cohere’s biggest upgrade yet — unifying enterprise‑grade security, intelligence, cost governance, and full‑stack control into a single agentic AI platform that runs wherever your data lives.”
“It is also notably our first Mixture of Experts (MoE) model with 25 billion active parameters ands 218 billion total parameters.”
“Our models run on any cloud AI/ML platform, including OCI, Azure, AWS, and Google CP”
How AI features are evaluated before and after release, evals, red-teaming, monitoring.
Assessed against NIST GenAI · NIST AI RMF · ISO 42001
“We simulate adversarial attacks on our systems to test for vulnerabilities and improve defenses”
“We carefully track the lineage of our training data to guard against data poisoning and malicious content We evaluate our models against internal and external benchmarks”
“Rigorous security: Proactively identify and mitigate potential threats through continual security testing, including red-teaming exercises and third-party vulnerability scanning.”
How changes to models and AI features are controlled and communicated: versioning, notice, rollback.
Assessed against ISO 27001 · ISO 42001
“Effective April 4, 2026, the following models are no longer available. Requests using these model IDs will fail.”
“System observability: Gain visibility into North's operations with easy-to-use tools for monitoring service performance and detailed logs of all changes.”
AI system description: vendor-evidenced, not yet independently corroborated.
Change management: vendor-evidenced, not yet independently corroborated.
Model10% of the score40
Which model providers and versions process customer data, and how changes to them are disclosed.
Assessed against NIST GenAI · ISO 42001
“Bring your organization’s existing models: North is model agnostic, use our generative model offerings or bring your favorites.”
“We may use our own AI technology or third-party AI tools to analyze the contents of support tickets. This helps us summarize inquiries, draft responses, and resolve issues faster. Please do not include sensitive, confidential, or personal information in your support tickets.”
“It is also notably our first Mixture of Experts (MoE) model with 25 billion active parameters ands 218 billion total parameters.”
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
“Cohere does not have access to data processed in connection with the use of Cohere models or products in Private Deployments or Deployments on Third-Party Managed AI services (e.g. on AWS, Azure, or GCP). The data remains within the enterprise customer’s or the third-party cloud provider’s controlled environment and is never accessible to or used by Cohere (for example, we do not train our models on it).”
“we collect content you submit to the Cohere Products, and outputs generated by the Cohere Products. We may use this input and output data to conduct research and development. This includes analyzing usage trends and using inputs/outputs to improve the performance and safety of Cohere Products.”
“However, as a further safeguard, Cohere takes steps to de-identify input and output data prior to any use for research and development purposes. If you need to include personal information in inputs to Cohere Products, consider moving to an enterprise tier and entering into a commercial agreement and DPA with Cohere.”
“Maintain full control over your data with customizable handling and retention settings. Opt out of model training at any time — your data stays yours.”
“Enterprise Users of the Platform can consult retention practices for inputs and outputs on the Platform here . Retention of inputs and outputs on the Platform is generally 30 days for Enterprise Users.”
“Examples of countries we transfer personal information to include, but are not limited to, Canada, the United States, and the United Kingdom. As a result, in certain circumstances, other foreign governments, courts, law enforcement agencies, or regulatory agencies may be entitled to access the personal information collected and held by Cohere, its affiliates, partners, and service providers.”
Customer data treatment: vendor-evidenced, not yet independently corroborated.
AI supply chain10% of the score40
The named chain of AI and infrastructure providers behind the product, and how it is disclosed.
Assessed against ISO 27001 · ISO 27017
“Our cloud infrastructure provided by GCP in the United States, for instance, is protected under the EU-US Data Privacy Framework , including its UK extension.”
“Users of the Platform or Cohere APIs can consult our list of sub-processors , which includes data processing locations.”
“Additionally, our collaboration with NVIDIA finds that North generates more insights, for fewer tokens, with Cohere models trained and deployed on NVIDIA Blackwell and Hopper.”
Security foundation15% of the score50
How security problems are found, reported and handled: disclosure channels, bug bounty, incident response.
Assessed against ISO 27001 · NIST AI RMF
“The Cohere Bug Bounty program is designed to recognize the community’s support in protecting our users’ privacy and security. To learn more about the program, download our Responsible Disclosure Policy .”
“Our advanced threat detection systems continuously monitor for suspicious activity”
“We conduct annual third-party audits and penetration tests to identify and address vulnerabilities”
Independent assurance evidence10% of the score41
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
“Our API platform is hosted on industry-leading cloud infrastructure and is SOC 2 Type II compliant”
“Exceeding global security standards: with SOC 2 Type 2, ISO 27001, and ISO 42001 certifications.”
“We conduct annual third-party audits and penetration tests to identify and address vulnerabilities”
None of the corroborated certifications is AI-specific (ISO/IEC 42001 or AIUC-class) — this assurance is security and data-transfer attestation, not AI-management-system assurance.
Legal & contractual10% of the score40
The contractual backbone a buyer can rely on, terms, DPAs, privacy commitments and regulatory posture.
Assessed against EU AI Act · OAIC APPs
“Cohere’s use of Customer Data is governed by the terms and conditions of the enterprise agreement and, where applicable, the Data Processing Addendum entered into between the parties.”
“Website Terms of Use Commercial SaaS Agreement Client License Agreement (CLA) Service Level Objective (SLO) Azure AI EULA for Cohere Models Channel Partner Marketplace Terms of Service”
“If we receive a request from an End User to exercise rights with respect to their personal information that we maintain on behalf of an enterprise customer, we will direct that individual to the relevant enterprise customer who is responsible to the End User for that request.”
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: 47 → up to 70 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 3, registry checks and verification ladders, click to view
Claimed for the Cohere API platform and for North; no auditor, report period or scope statement in the collected documents.
Not yet checked against a registry. Treat it as the vendor’s own claim and ask the vendor for the current certificate.
Named only in the North 2 announcement; certification body, scope and validity not stated.
Not yet checked against a registry. Treat it as the vendor’s own claim and ask the vendor for the current certificate.
Named only in the North 2 announcement; certification body, scope and validity not stated.
Not yet checked against a registry. Treat it as the vendor’s own claim and ask the vendor for the current certificate.
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 Cohere 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.