Enterprise-Grade Security, Built Into How We Deploy
Conversational BI that meets PIPL, GDPR, PDPO and the EU AI Act — with data residency, IM-native security inheritance, and AI governance by default.
Compliance Across Jurisdictions
We operate under overlapping regimes so your deployment is defensible wherever your teams sit.
PIPL · DSL · CSL
China's personal information, data security, and cybersecurity laws — with MLPS 2.0 Level 3 in progress.
PDPO & CIO Ordinance
Hong Kong's Personal Data Privacy Ordinance and the new Critical Information Infrastructure regime.
GDPR & EU AI Act
EU data protection plus the risk-tiered AI Act obligations for high-impact systems.
PDPA · CCPA/CPRA
Singapore and US state privacy frameworks, covered for cross-border enterprise clients.
Fifteen Years of Delivering Enterprise Systems
Long before conversational AI, our team shipped secure, regulated systems for institutions that cannot afford to be wrong.
Foundation
Enterprise data platforms and integration for financial and retail clients.
Cloud & scale
Multi-region, compliant cloud architectures with audited access controls.
AI in production
First conversational and agentic deployments inside client IM environments.
Forward-deployed
Governed AI embedded where teams already work, with security inherited from the platform.
Encryption & Isolation by Default
Every deployment starts from a zero-trust baseline, not an enterprise upsell.
Encrypt everything
All data is encrypted in transit with TLS 1.3 and at rest with AES-256.
Per-tenant isolation
Client data is fully isolated; one client's data is never visible to another.
Role-based access
Least-privilege access with complete audit logging on every action.
No raw data to LLMs
Raw client data never leaves the secure environment and is never sent to model prompts.
Aggregated only
Only aggregated, anonymised query results are sent to enterprise-grade LLM providers.
Human review option
Enterprise clients can require human review before any AI answer is delivered.
We Never Train on Your Data
Our governance model is the product. It is not a checkbox we add at the end.
No model training
Client data is never used to train any LLM. Models are general-purpose and pre-trained.
Prompt boundary
Only aggregated, anonymised results reach the model for natural-language generation.
Auditability
Every query, answer, and data access is logged and reviewable on request.
Raw rows never cross the boundary. The model sees a summarised result, never the underlying records — and enterprise clients can require a human checkpoint before an answer is delivered.
Your Data Stays Where It Belongs
China client data stays in mainland China; Hong Kong data stays in Hong Kong. Only aggregated or anonymised results may cross borders.
China SCC / GBA / CAC
Where personal data must cross borders, we use China SCC, the GBA Standard Contract, or a CAC Security Assessment.
EU SCCs + TIA
For EU data we use EU Standard Contractual Clauses with Transfer Impact Assessments.
Your IM Platform's Security Is Ours
We deploy inside the platform you already secured. Each channel inherits enterprise-grade security — we add scoped permissions and data minimisation on top.
| Channel | Platform security | Our additional layer |
|---|---|---|
| WeChat Work | Enterprise authentication, Tencent cloud security | ISV-level data isolation, no credential storage |
| DingTalk | Alibaba cloud security, enterprise SSO | ISV sandbox, data minimisation |
| Feishu | ByteDance enterprise security | Bot API security, scoped permissions |
| End-to-end encryption, Meta enterprise security | Token-based API access, no message content stored | |
| Microsoft Teams | M365 security, Azure AD, DLP policies | Bot Framework security, tenant isolation |
| Telegram | MTProto encryption, bot token security | No sensitive data in messages, auto-delete timers |
| Slack | Enterprise Grid security, SSO, DLP | App-level permissions, scoped tokens |
We don't add a new attack surface. We work within the security framework you've already secured, and every deployment inherits the enterprise-grade security of the platform you choose.
Targeting Industry Standards
We are transparent about where we are and where we are heading. Our architecture already meets these standards — formal certification is the next step.
ISO 27001:2022
Global standard for information security management. 93 controls across 4 themes.
MLPS 2.0 Level 3
China's de facto benchmark for serious B2B SaaS — management systems, segmentation, access control, encryption, incident response.
SOC 2 Type II
US/international SaaS standard covering security, availability, processing integrity, confidentiality, and privacy.
Our security architecture already meets these standards. We encrypt all data in transit and at rest, isolate each client's data, never expose raw data to LLMs, and operate within your existing IM security framework across WeChat Work, DingTalk, Feishu, WhatsApp, Microsoft Teams, and Telegram.
A Six-Phase Response Protocol
When an incident occurs, speed and transparency matter. Our protocol defines containment, communication, and remediation.
Detection
Automated monitoring plus manual reporting.
Assessment
Classify severity and determine scope.
Containment
Isolate systems and revoke access.
Notification
Notify clients and regulators.
Remediation
Fix the vulnerability and restore service.
Post-incident
Root-cause analysis and prevention.
Security FAQ
Practical answers about our security practices, compliance posture, and data governance.
Beehive Strategy encrypts all data in transit using TLS 1.3 and at rest using AES-256. Client data is fully isolated per tenant, with role-based access control and complete audit logging. Raw client data never leaves the secure processing environment and is never sent to LLM prompts. Only aggregated, anonymised query results are sent to enterprise-grade LLM providers for natural language generation.
In China, we comply with PIPL, DSL, CSL, and are targeting MLPS 2.0 Level 3 certification. In Hong Kong, we comply with PDPO and the CIO Ordinance. Internationally, we comply with GDPR, the EU AI Act, Singapore PDPA, and applicable US state privacy laws including CCPA/CPRA.
No. We never use client data to train any LLM models. Only aggregated and anonymised query results are sent to the LLM for natural language generation. Raw client data never leaves the secure processing environment. Enterprise clients can additionally require human review before any AI-generated answers are delivered.
China client data stays in mainland China infrastructure, and Hong Kong client data stays in Hong Kong infrastructure. Only aggregated or anonymised results may cross borders, which may qualify for China's Scenario A exemption. Where personal data must cross borders, we use China SCC, GBA Standard Contract, or CAC Security Assessment for China transfers, and EU SCCs with Transfer Impact Assessments for EU transfers.
We are pursuing ISO 27001:2022 (targeting 6–12 months), MLPS 2.0 Level 3, and SOC 2 Type II (targeting 6–18 months). Our security architecture already meets these standards — we encrypt all data in transit and at rest, isolate each client's data, never expose raw data to LLMs, and operate within existing IM security frameworks.
Ready to Deploy AI Your Security Team Will Approve?
Start with a free readiness assessment, or book a working session to walk through your data-residency and governance requirements.