AI RISK ASSESSMENT × INSURANCE
Make AI Risk Insurable
AgentSure quantifies invisible AI risk and arranges the cover to transfer it — we assess and price; licensed insurers carry the risk and pay the claims.
QUANTIFY
MITIGATE
TRANSFER
Three ServicesWHAT WE DELIVER
01
QUANTIFYQuantifymake AI risk visible
METHODWhite-box + black-box testing for AI-specific risks (privilege escalation, prompt injection, tool poisoning, hallucination), distilled into underwriting signals.
YOU GETYour AI's risk exposures and severity levels — structured, underwriting-ready.
02
MITIGATEMitigatemake AI risk fixable
METHODPer-exposure fixes (guardrails, prompt rewrites, tool limits, fail-safes), re-tested to confirm the risk has dropped; high-severity items remediated jointly with CRPO ecosystem vendors.
YOU GETAn itemised remediation plan + a re-test report (better posture, lower premium).
03
TRANSFERTransfermake AI risk insurable
METHODAsian reinsurers backstop residual risk; quantification drives pricing; at claim time the insurer pays out — never the client.
YOU GETCoverage + an “AI go-live clearance” Legal, IT, Risk & Management can sign.
Four-Layer SystemR&D → COVERAGE
Four interlocking layers, from R&D to insurance value — we map the whole Agent supply chain (every tool your agents call), not just one system.
①Insurance ValueRisk-priced cover · go-live clearance · lower risk, lower premium · mitigation built in
②Full-Scan · CoverageMCPAgents (A2A)NetworkSkillsPaymentsone risk model, ported across
③Risk Engine · Capability50-class risk map · 100+ detection rules · black-box testing + white-box code analysis
④Continuous R&D · FoundationOngoing R&D with CRPO / SMU / NTU
Why Trust UsPEDIGREE
Cyber SecurityCSAR&D ProgrammeCRPOAI-Risk QuantCore ProjectLicensed IPSpin-outAGENTSURE
SMU Spin-outNational Standard
TR 99:2021 Co-author Team
Among the lead authors of Singapore's AI-security national standard; that methodology is distilled into our risk engine.
Government Track Record
High-Security Deployments
Research deployed in high-security government settings, outperforming off-the-shelf tools.
Actuarial × Insurance
Tiger Li · CEOYonghu Li
FSA · FCAA · CERA — 18+ yrs across actuarial, consulting, reinsurance & InsurTech.
AI Security
Xiaofei Xie · Chief ScientistXie Xiaofei
SMU Professor · 6× ACM SIGSOFT Distinguished Paper Awards · 150+ top-venue papers.
Next step — book an AI risk assessmentwww.agentsure.tech →
AI 风险测评 × 保险
让 AI 风险 真正可保
AgentSure 测评并量化不可见的 AI 风险,再对接保险把它转移出去——我们出数据与定价,持牌保险公司与再保人承保并理赔。
QUANTIFY量化风险
MITIGATE缓释风险
TRANSFER保险兜底
三大服务THREE SERVICES
01
QUANTIFY量化让 AI 风险可见
方法白盒代码审查 + 黑盒系统测试,专测越权、注入、投毒、幻觉等 AI 特异风险,提炼核保信号。
你会拿到AI 的风险暴露和对应的风险水平(结构化、能辅助核保)
02
MITIGATE缓释让 AI 风险可修
方法逐个风险暴露点出修复方案(护栏、提示词、工具边界、降级熔断),重测确认下降;高风险可对接 CRPO 安全生态厂商协同修复。
你会拿到逐条修复方案 + 重测确认报告(做得越好,保费越低)
03
TRANSFER保险让 AI 风险可保
方法联合亚洲再保人,把残余风险兜底,量化结果作为定价依据;出险后保险公司理赔,不甩单给客户。
你会拿到承保方案 + 一份让法务、IT、风控、管理层点头的「AI 上线许可」
四层体系 · 持续领先FOUR-LAYER SYSTEM
从国家级研发一路贯通到保险价值,四层咬合、层层向上 —— 我们不止测一个系统,而是持续测绘整个 Agent 供应链(你的 Agent 调用的全部外部工具),在 AI 风险上持续领先。
①保险价值层按风险定价的保险 · 上线许可 · 风险越低、保费越低 · 内置缓释
②全量扫描 · 覆盖层MCP 端Agent 端 (A2A)网络端Skill 端支付端同一套风险模型平移
③风险引擎 · 能力层50 类风险地图 · 100+ 条检测规则 · 黑盒测试 + 白盒代码分析引擎
④持续研发 · 根基层CRPO / SMU / NTU 持续研发
为什么信得过PEDIGREE
网络安全局CSA研发计划办公室CRPOAI 风险量化核心子项目IP 正式授权Spin-outAGENTSURE
SMU Spin-out国家标准
TR 99:2021 主笔团队
AI 安全国家标准主笔之一,方法论沉淀为风险量化引擎。
政府场景
高安全场景实践
研究成果已应用于政府高安全场景,效果超越市场常见安全工具。
精算保险
Tiger Li · CEO李用虎
FSA · FCAA · CERA —— 18+ 年精算 / 咨询 / 再保 / InsurTech。
AI 安全
Xiaofei Xie · 首席科学家谢肖飞
SMU 教授 · 6× ACM SIGSOFT 杰出论文 · 150+ 顶会论文。
下一步 —— 预约一次 AI 风险评估www.agentsure.tech →