For a precision manufacturer, the CAD model is the product. It holds the geometry, tolerances and design intent that took years of engineering to develop. In aerospace and defense, it may also be export-controlled under ITAR or classified as Controlled Unclassified Information (CUI). Losing control of that file is not an IT incident. It is a business, contractual and sometimes legal one.
That is why the first question an IT or information security team asks about any AI CAM software is not "how fast is it?" It is "where does our data go?"
AI tools raise that question more sharply than traditional CAM. Many AI products send data to shared cloud models, retain prompts for training, or route requests through third-party model providers. For a manufacturer handling defense programs or proprietary designs, any of those can rule a vendor out before the evaluation starts.
Six questions to ask any AI CAM vendor
Before approving an AI CAM tool, security teams should get clear, written answers to these questions:
- 1
What data leaves the workstation?The whole assembly, the open part only, or metadata? And does the tool reach into PLM or tool management systems?
- 2
Where is it processed and stored?Which cloud, which region, and can you choose your own region for data residency?
- 3
Is our data used to train models?Including models that serve other customers.
- 4
Which AI models are used, and who hosts them?Are prompts and outputs shared with an external model provider?
- 5
How is data isolated from other customers?Logical separation, dedicated storage, customer-managed keys?
- 6
Can it support ITAR and CUI workloads?Is there a deployment where infrastructure access is restricted to U.S. persons?
If a vendor cannot answer all six clearly, that is the answer.
How the Limitless Labs CAM Agent protects CAD data
The Limitless Labs CAM Agent works inside Siemens NX, Mastercam and PTC Creo. It was designed for manufacturers whose CAD data is their most sensitive asset, including aerospace and defense programs. Six commitments define how it handles that data.
Encrypted in transit and at rest.Every connection uses TLS 1.2 or higher. Data at rest is encrypted with AES-256, with keys managed in AWS Key Management Service (KMS).
Isolated per customer.Storage, inference and logs are segregated by customer ID. Customer-managed keys add physical separation for organizations that require it.
Not used to train shared models.Your data does not train the general model. Any customization built on your tool library or machining standards serves your organization only.
No third-party AI access.The large language models behind the agent are hosted privately on Amazon Bedrock in AWS GovCloud (US). Prompts and outputs are never shared with model providers.
Limited to the CAM session.The agent reads the open part and the session context. Optional connections to tool management (TMS) and PLM systems are read-only.
Deleted on request.Retention periods are set in your agreement, and data is erased in line with NIST SP 800-88 media sanitization guidelines.
Certifications and frameworks
| Framework | Status |
|---|---|
 ISO/IEC 27001:2022 | Certified |
 SOC 2 Type II | Audited, report available under NDA |
| AWS GovCloud (US) | Available for ITAR and CUI workloads |
 FedRAMP High | Authorization in progress, expected January 2027 |
| GDPR and CCPA | Compliant |
| FIPS 140-2 and CIS benchmarks | Infrastructure hardened to both |
Deployment options: from commercial to ITAR
Not every manufacturer needs the same level of isolation. The CAM Agent is available in three deployment models:
- •Limitless Labs Cloud (standard). Hosted on AWS for commercial and industrial customers, with complete logical segregation by customer ID.
- •Enterprise: your keys, your region. Customer-managed keys with dedicated storage and data residency in an agreed AWS region, plus SSO and SIEM integration.
- •ITAR and CUI: AWS GovCloud (US). Infrastructure access is restricted to U.S. persons, which supports ITAR compliance requirements for export-controlled technical data. FedRAMP High authorization is expected in January 2027.
Security controls reviewed in vendor assessments
These are the controls security teams most often check in a vendor security questionnaire:
| Area | Controls |
|---|---|
| Access | SAML 2.0 SSO with Azure AD, Okta and other identity providers. Enforceable MFA, session timeouts and login lockout. Role-based access, with internal staff access on a need-to-know basis. |
| Encryption | TLS 1.2 and 1.3 with FIPS 140-2 hardened cipher suites. AES-256 at disk, application and field level. AWS KMS keys rotated at least annually, with every key event logged. |
| Monitoring | Continuous monitoring with AWS GuardDuty, Security Hub and Inspector. Admin and access logs kept for 6+ months, read-only and exportable. SIEM integration available by agreement. |
| Secure development | Aligned with the OWASP Top 10, with peer review and static and dynamic scanning. Independent penetration test at least once a year. CIS-hardened systems with separate development, staging and production environments. |
| Resilience | Authorities notified within 72 hours of a privacy breach, and affected customers informed directly. Daily encrypted, immutable backups in a separate region or zone. RPO 24 hours, RTO 2 days, with disaster recovery tested annually. |
| Retention | Retention and termination terms defined in your agreement. Secure erasure on request in line with NIST SP 800-88. Backups expire on a fixed schedule: daily backups are kept 30 days, weekly 26 weeks, monthly 12 months and yearly 7 years. |
Documentation available for your review
Security teams can request the ISO/IEC 27001 certificate, the SOC 2 Type II report and penetration test report (both under NDA), the subprocessor list, and completed security questionnaires. An architecture review with the Limitless Labs team is also available. Start at the Limitless Labs Trust Center or contact contact-us@limitless-labs.ai.
Frequently asked questions
Does the Limitless Labs CAM Agent use our CAD data to train AI models?
No. Customer data does not train the general model. Any customization built from your library serves your organization only.
Is the CAM Agent ITAR compliant?
For ITAR and CUI workloads, the CAM Agent runs on AWS GovCloud (US), where infrastructure access is restricted to U.S. persons. This deployment is designed to support ITAR compliance requirements for export-controlled technical data.
Are our prompts or parts shared with OpenAI, Anthropic or other model providers?
No. The models are hosted privately on Amazon Bedrock in AWS GovCloud (US), and prompts and outputs are never shared with model providers.
Can we keep our data in a specific region and control the encryption keys?
Yes. The Enterprise deployment offers customer-managed keys, dedicated storage and data residency in an agreed AWS region.
What data does the agent access?
The open part and the CAM session context. Connections to PLM and tool management systems are optional and read-only.
Is Limitless Labs FedRAMP High authorized?
FedRAMP High authorization is in progress and expected in January 2027.
Bring your security team into the evaluation early.
Request the Limitless Labs security documentation or book an architecture review.