Shadow AI & secret leakage影子 AI 与数据外泄
Confidential data reaches public copilots before security teams see it.敏感数据流向公有助手，安全团队未能及时感知。
AI Governance PlatformAI 安全合规平台
Govern Every AI Interaction.AI 安全、合规和隐私防火墙
Review every AI interaction before it reaches external AI models. Block unsafe behavior, protect sensitive data, and attach audit evidence on every request.在流量触达外部 AI 模型之前完成审查。拦截风险、保护敏感数据，并为每次请求留存可审计证据。
Enable businesses and individuals to safely access external AI models while preserving security, compliance, and privacy. 帮助企业和个人安全调用外部模型，守住安全、合规与隐私底线。
- Review before external AI外部调用前审查
- Least-privilege agent controls智能体最小权限控制
- Cross-border compliance built in内置跨境合规能力
- Audit evidence per request每次请求审计证据
Three control gaps across your AI stack.AI 架构的三类管控盲区。
Privacy, AI security, and data compliance in one runtime loop across gateway, cloud APIs, and plugins—with evidence on every decision.隐私、安全与合规在同一运行时闭环内联动，覆盖网关、云端 API 与插件，每次决策均可追溯。
Runtime governance loop运行时治理闭环
Policy-trained small models across your AI stack connect every stage, from traffic capture to audit-ready evidence.策略内置小模型覆盖全栈，串联从流量捕获到可审计证据的完整链路。
One platform. Three modules. Two deployment paths.一个平台、三个模块、两种部署形态。
Policy Engine trains formal policy language into models. AI Governance enforces at the traffic layer—Firewall or Guard. Add Cross-Border when transfer compliance is required.Policy Engine 将形式化制度语言训入模型；AI Governance 在流量层执行——Firewall 或 Guard 两种形态；按需叠加 跨境合规。
Policy Engine
Underlying technology底层技术
Trains
regulations, corporate bylaws, and industry
standards
into proprietary small models.将法规、企业条例与行业规范训入自研小模型。
Learn more了解详情 →
AI Governance
Two deployment paths两种部署形态
Sovereign AI Firewall or
AI Traffic Guard—runtime enforcement across
your AI stack.Sovereign AI Firewall 或
AI Traffic Guard——覆盖 AI
全栈的运行时治理。
Learn more了解详情 →
Cross-Border跨境合规
Transfer compliance传输合规
Per-transfer jurisdiction checks. Standalone
or integrated with AI Governance.每次传输辖区审查，可独立部署或与 AI
Governance 集成。
Learn more了解详情 →
Why traffic-level governance is different.为什么流量级治理不同。
Act while risk is still controllable: before transmission, before agents execute, and before evidence is lost.在风险仍可处置时行动：传输前、智能体执行前、证据丢失前。
Approach治理方式
What it usually achieves通常实现什么
Mode IO differenceMode IO 的差异
Output moderation输出审核
Checks responses after the request is sent.请求发出后才检查模型回答。
Control prompts, files, agent actions, and APIs before risk
leaves.风险离域前控制提示词、文件、智能体与接口。
Traditional DLP传统数据防泄漏
Finds static patterns; misses intent, context, and
agents.识别静态模式，遗漏意图、上下文与智能体行为。
Small models for privacy, compliance, and AI safety
context.专用小模型理解隐私、合规与 AI 安全上下文。
Cloud-only compliance tools纯云端合规工具
May require sensitive traffic to leave first.可能要求敏感流量先离开企业。
Local-first and hybrid deployment at the traffic
source.本地优先与混合部署，处置靠近流量源。
Research-grade models. Enterprise-grade controls.研究级模型，企业级控制。
Privacy, compliance, and AI safety—enforced in real time across your AI stack.在您的 AI 环境中实时落实隐私、合规与安全管控。
Stop unsafe agent actions before they run.在智能体执行前拦截不安全操作。
Detect prompt hijacking, tool abuse, and risky execution at every AI exit point.在每一个 AI 出口识别提示词劫持、工具滥用与高风险执行。
Protect sensitive data without stopping work.保护敏感数据，不中断业务。
Identify PII, trade secrets, and regulated data in context; risk-appropriate handling at request time so work keeps moving.识别个人、商业机密与受监管数据；按风险等级匹配处置，业务不中断。
Dynamic baselines by identity, scenario, and purpose.按身份、场景与目的动态调整合规基准。
Apply GDPR, CCPA, HIPAA, EU AI Act, EU Data Act, PIPL, and more in real time with a full auditable evidence chain.实时适用 GDPR、CCPA、HIPAA、欧盟 AI 法案、欧盟数据法、PIPL 等法规，留存完整可审计证据链。
GDPRCCPAHIPAAEU AI ActEU Data ActPIPLetc.
Customer scenarios客户场景
From industry pain to从行业痛点到 governed AI可治理的 AI.。
AI CompaniesAI 企业
Pain: Over-scoped agents and hijacked prompts exfiltrate data before security teams see it.痛点：智能体权限过大或提示词被劫持时，数据可能外泄而团队难以及时察觉。
Mode IOLeast-privilege enforcement, ASI-class blocking, and P0–P3 audit trail.最小权限执行、ASI 类拦截与 P0–P3 审计留痕。
Learn more了解详情FinTech & Consulting金融科技 / 咨询
Pain: Deal terms and client emails reach public copilots unnoticed.痛点：交易条款与客户邮件悄然进入公共协同助手。
Mode IODetect trade secrets in context; handled according to risk level, with every action logged.识别商业机密，按风险分级处置，全程留痕。
Learn more了解详情Healthcare医疗健康
Pain: PHI in copilot prompts leaves the boundary with no HIPAA review.痛点：健康信息随协同助手提示词出境，缺乏 HIPAA 审查。
Mode IORisk-appropriate handling before transmission; retain auditable evidence packets.传输前按风险等级匹配动作处理 PHI，留存可审计证据包。
Learn more了解详情Cross-Border Trade跨境贸易
Pain: Employee records reach foreign LLMs with no proof under regulations like PIPL and GDPR.痛点：员工记录流向境外大模型，缺乏 PIPL、GDPR 等多项法规下的合规证据。
Mode IOLive jurisdiction and purpose checks. Risk-appropriate enforcement with audit evidence.实时辖区与目的审查，按风险分级处置，并留存审计证据。
Learn more了解详情