World-Class Agentic Security Expert
Jeremy Canale I design ▍
Twelve years securing the world's most regulated industries, banking, insurance, defense. Now architecting the security perimeter for autonomous AI agents, MCP infrastructure, and the cloud platforms they run on. Also working on AI safety, alignment and the oversight of AI agents, the subjects of my latest research papers on arXiv.
Hands-on experience for
- Swiss Re
- AXA
- BNP Paribas
- Standard Chartered
- Thomson Reuters
- Société Générale
- Seddiqi Holding
- NusaVest
- Hal · SC Ventures
- Thales
- Dassault Systèmes
- Veolia
- Baxter
- Gemalto
Credentials
Certified across the agentic stack
From Anthropic AI Fluency to NVIDIA Generative-AI LLMs, ISACA CISM/CRISC, and the full Microsoft Cybersecurity Architect & AWS Security tracks.
Building Secure Agents for Public Sector Services
Agentic automation security curriculum.
Certified Professional · Agentic AI
Certified Associate · Generative AI LLMs
AI Fluency Framework & Foundations
ud6tp3nbypsc
Claude Code in Action
bqps2dqo8oaa
Introduction to Claude Cowork
qn62hbyfx7ky
Claude Code 101
cyrei883psj9
Claude 101
pok9weot8j3p
Prisma Certified Cloud Security Engineer
Cybersecurity Architect Expert
EAB47D8F4EDC1058
DevOps Engineer Expert
D7EF2C23EF149EA4
Azure Developer Associate
A6B7126C06F9C786
Security Operations Analyst Associate
8B08474D2BE410E8
Security, Compliance & Identity Fundamentals
BD23427E7A5DF944
Azure Fundamentals
B7635962A1631A02
Azure Security Engineer Associate
H344-1450
CISM, Certified Information Security Manager
CRISC, Risk & Information Systems Control
AWS Certified Security, Specialty
6S5MRC92DFR11RCW
AWS Certified SysOps Administrator, Associate
FZN3QCE2J21E1N5F
AWS Certified Cloud Practitioner
WH8FRT11JNBQQXCM
AWS Certified Developer, Associate
4NBMCXH12EVEQB36
AWS Certified Solutions Architect, Associate
AWS-ASA-14402 · 3Z8KF6HK22VQQ83R
Managing Security in Google Cloud Platform
Google Cloud Platform Fundamentals · Core Infrastructure
Professional Scrum Master I
Cisco Certified Network Associate (CCNA)
Domains
Where I operate at the frontier
From securing autonomous AI agents to underwriting cyber risk for global insurers, my work sits at the intersection of regulated industries and emerging AI.
Agentic AI Security
Securing autonomous LLM agents end-to-end: MCP servers, tool-call authorization, prompt-injection defense, sub-agent isolation, and runtime guardrails.
Signature practiceAI Governance & Trust
Operationalizing ISO/IEC 42001, NIST AI RMF and the EU AI Act. Bringing Anthropic's AI Fluency Framework into the enterprise risk register.
Cloud Security & CSPM
Multi-cloud security architecture for AWS, Azure and GCP. Global Prisma Cloud deployment at AXA. Cloudflare estate at Thomson Reuters. Hardened blueprints, IaC and SRE practices.
Cyber Risk Quantification
Pricing, exposure and portfolio analytics for insurers and reinsurers.
GRC & Compliance
ISO/IEC 27001 group certifications, SOC 2, HIPAA, DORA, NIS2, UAE & KSA PDPL. Built the GRC program at Seddiqi Holding and the BNP Paribas maturity roadmap.
Threat Intel & Offensive
Published OWASP research in Hakin9. Pentesting, exploit prediction and attack-surface intelligence.
Trust by design
One architect. Every framework that matters.
From ISO/IEC 27001 to the EU AI Act, my engagements bridge engineering and regulation. I translate frameworks into deployable controls, and prove them with measurable assurance.
- Cyber maturity 0 → 30% in <12 months for BNP Paribas Asset Management.
- Global Prisma CSPM rollout across all AXA entities, worldwide.
- Group ISO 27001 program at Seddiqi Holding, aligned to UAE & KSA PDPL.
- Security architecture gatekeeper at Swiss Re & Thomson Reuters.
Career
Twelve years at the security helm of regulated enterprises
Hands-on roles, from Vice-President at Swiss Re to Deputy CISO at the BNP Paribas Asset Management group.
-
Jun → Sep 2026
Fractional CISO · Hal · SC Ventures × Microsoft
Security leadership for Hal, SC Ventures' born-in-AI, agentic-first growth platform for GCC SMEs, built with Microsoft. Cyber strategy, ISMS, cloud and AI-agent security governance. Mission ended in September 2026.
-
2026
Fractional CISO · NusaVest
Cyber strategy and ISMS for a Southeast-Asia asset-backed securities platform (SC Ventures · Standard Chartered).
-
2025 → 2026
AI GRC Lead · Seddiqi Holding
Group ISO/IEC 27001 certification program across the holding. Risk register, Statement of Applicability, UAE & KSA PDPL alignment for every department.
-
2023 → 2025
Deputy CISO · GAMBIT / BNP Paribas Asset Management
Built a NIST-based cybersecurity program from the ground up. After BNP Paribas acquisition, drove the maturity transformation from 0% to 30% within the group framework.
-
2023 → 2024
Head of Cloud Security GRC & CSPM · AXA
Technical lead for the worldwide rollout of Palo Alto Prisma CSPM across every AXA entity. Aligned global standards with local regulatory constraints.
-
2022 → 2023
Cloud & Product Security Lead · Thomson Reuters
Security architecture gatekeeper within enterprise architecture governance. Global Cloudflare deployment. Cyber-underwriting advisory bridging engineering and insurance.
-
2020 → 2022
Vice-President · Cloud Security & GRC · Swiss Re
Security by design across the enterprise. Architecture review for every major program. Reusable security blueprints across cloud, application and infrastructure domains.
-
2018 → 2019
AI CyberSecurity Architect · Société Générale
Defined and enforced the cloud security framework for the bank. Risk-based cloud governance and security-by-design for every cloud initiative.
-
2015 → 2018
Security Solutions Architect · multiple Fortune-500
Cloud migrations under PCI-DSS, HIPAA and regulated workloads. IaC, hardened blueprints, vulnerability automation, native cloud SSO.
Consulting
Hands-on consulting for AI-native enterprises
End-to-end advisory and implementation across the agentic security stack, from boardroom strategy and architecture blueprints to production-grade guardrails on Azure AI Foundry, LangGraph, NeMo, SageMaker and Prisma Cloud.
Phase 01 · Strategy
Agentic AI Security Strategy
Boardroom-level diagnostic of your AI agent estate. Target architecture, regulatory posture, and the 12-month roadmap to get there.
- Threat model of every agent & tool call
- Mapping to NIST AI RMF, ISO 42001, EU AI Act
- CISO-ready risk & cost model
Phase 02 · Build
Build & Harden
Forward-deployed engineering on your stack. Guardrails, sub-agent isolation, tool authorization, Langfuse observability, shipped to production.
- Reference architecture on Azure AI Foundry / AWS Bedrock
- LangGraph + LangChain workflows with guardrails
- Prisma Cloud + Cloudflare integration
Phase 03 · Assure
Audit & Assurance
Independent agentic security audit. Red-team prompt injection, tool-chain abuse, data exfiltration. Boardroom-ready evidence pack.
- Red-team across prompt, tool & data layers
- SOC 2 / ISO 27001 / DORA / NIS2 mapping
- Continuous-assurance Langfuse dashboards
01
Agentic AI platforms & orchestration
02
Model serving & inference
03
Cloud security & CSPM
04
Observability, evals & guardrails
05
Governance, risk & compliance
Ready to secure your agentic platform?
Initial scoping call, typically 30 minutes, no commitment.
contact@jeremycanale.com
Case Studies
Selected engagements at the frontier of agentic security
Anonymised summaries of recent missions across cyber insurance, multi-cloud security and group GRC. Details vary by industry and are kept deliberately broad to respect client confidentiality.
Cyber Insurance
AI driven cyber underwriting platform
Challenge
Insurers and reinsurers needed real time, evidence based cyber risk data to replace static questionnaires and accelerate the pricing workflow.
Approach
Designed and shipped a multi OS desktop application backed by an AI rating engine, with MCP native distribution to common assistant clients and an underlying incident intelligence corpus.
Outcome
A production grade platform recognised across the cyber insurance market.
Global Insurance Group
Worldwide CSPM rollout
Challenge
A multinational insurer operating across dozens of entities had a heterogeneous cloud security posture and no unified visibility.
Approach
Technical lead for the global deployment. Cross entity workshops, alignment of group security standards with local regulatory constraints and translation into implementable controls.
Outcome
A single CSPM signal across the entire group, with regulatory alignment per jurisdiction.
Asset Management, Tier 1 Banking
Post acquisition cyber maturity
Challenge
A newly acquired entity needed a formalised cyber programme aligned to the acquirer's group framework, with measurable, defensible evidence of progress.
Approach
Built a NIST based cybersecurity programme from the ground up. Controls, policies and procedures, risk register, structured reporting to group level stakeholders.
Outcome
Validated maturity progress within the acquirer's framework.
Diversified Holding, GCC
Group ISO 27001 certification
Challenge
A multi business holding with no unified information security management system, operating under regional data protection laws.
Approach
High level designs, application level IT risk assessments, group policy framework, Statement of Applicability, and alignment to regional regulatory requirements.
Outcome
Group wide ISO 27001 readiness with structured assurance per department.
Global Financial Information Provider
Security architecture stewardship
Challenge
A global enterprise required consistent security by design across every major digital initiative.
Approach
Acted as security architecture gatekeeper within enterprise architecture governance, with reusable blueprints across cloud, application and infrastructure domains.
Outcome
Accelerated secure delivery and stronger, more consistent control maturity.
Southeast Asia, Asset Backed Securities
Fractional CISO mandate
Challenge
An emerging investment platform needed senior cyber leadership to define its information security strategy and ISMS from day one.
Approach
Fractional CISO engagement covering target operating model, risk appetite, control catalogue and the regulatory roadmap, in coordination with a global banking parent.
Outcome
A defensible cyber posture, ready for investor and regulator scrutiny.
Contributions
Public repositories and merged pull requests
My research repositories, the frameworks that cite them, and security fixes proposed to open source projects and merged by their maintainers. Pull requests are read from the GitHub API and only accepted ones appear; work still under review does not.
-
jeremy1392/RCAreplayPublic repository
ReplayRCA: diagnosing agent failures by counterfactual replay
Author's repository · code of the paper arXiv:2610.41025, co-authored with Fabien SoulisCan a model find the decisive mistake in another agent's failed run and fix it with a single action, after the fact or live? The repository holds the OpsWorld environment, the harvesting and calibration pipeline, the autopsy and live oversight tracks, the OpenRouter client and the reporting scripts. Each diagnosis is scored by restoring the environment before the chosen step, replaying the proposed action and comparing the outcome with a placebo.
Read the paper: arXiv:2610.41025
· MIT license · Python · 13 diagnosers · 62 calibrated failures · 87 supervised runs -
CyberStrategyInstitute/ai-safe2-frameworkCites my work
AI SAFE² cites DGF-Bench as a candidate benchmark for seven controls
Cyber Strategy Institute · open-source governance framework for agentic AI, v3.1, 161 controlsThe framework's technology contribution review maps DGF-Bench to seven controls across four of its five pillars, from Malicious Prompt Filtering to Multi-Agent Approval, and proposes reproducing its paired clean and attacked decisions. Nine works are reviewed in total; none has been validated or certified at this stage.
Read the article: DGF-Bench makes its way into AI SAFE²
· research/technology-contribution-examples.md · 7 controls · 4 of 5 pillars -
jeremy1392/DGF-BenchPublic repository
DGF-Bench: a defensive-security benchmark for agents acting as governance reviewers
Author's repository · code and package of the paper arXiv:2609.34913Each agent reviews a synthetic project dossier at a governance gate, with enterprise tools to read evidence and request what is missing. The same dossiers run in three conditions: structured facts, documents and records only, and attacked copies with manipulations in the evidence the organization does not vouch for. An attack counts only when the agent takes the exact injected action and does not take it on the paired clean run.
Cited by the AI SAFE² framework as a candidate benchmark: read the article
· Release v0.1.2 · 61 executable rules · 6 models · paired clean and attacked runs -
SalesforceAIResearch/BeagleMerged
fix(security): verify local Docker removal before confirming cleanup #3
Salesforce AI Research · rollout engine for evaluating and evolving AI agentsThe local Docker runtime cleared its container handle before stopping and removing the container, and ignored failed exit codes. A failed removal could leave an agent workload running while later cleanup calls did nothing. The fix forces removal even when the stop fails or times out, confirms the container is gone with an independent Docker query, and only then clears the handle. Otherwise it raises an error and keeps the handle for a retry.
Motivated by the containment questions raised by the OpenAI / Hugging Face incident: stopping an agent workload needs a verifiable outcome.
· 4 files · +288 -19 · 15 new regression testsView the pull request
Publications & Recognition
A decade contributing to the public conversation
OWASP vulnerability research and an early social media search engine recognised at the highest level.
-
2013Magazine, OWASP research
Hakin9 Magazine, international IT security publication
In depth research on OWASP based vulnerabilities affecting major platforms, published in a renowned international IT security magazine.
-
2012Product, recognition
AnoSearch, social media search engine
Built an early social media search engine extracting signal from more than twenty networks. Recognised as a top national IT project of the year and presented to top CEOs and to the French Minister of Defense.
-
2024Industry press
Commentary in the international reinsurance press
Featured commentary on AI driven cyber underwriting and data driven cyber insurance in leading international reinsurance media.
Manifesto
Autonomous agents need a new perimeter.
Legacy security was built for humans clicking buttons. Agents click thousands per minute, call external tools, spawn sub-agents and chain decisions across systems. The blast radius is no longer a session, it's a workflow.
My work re-anchors the perimeter at the place where intent meets execution: the tool call. I design authorization, observability and policy guardrails that make autonomous agents auditable, reversible and trustworthy at enterprise scale.
"In the agentic era, the question isn't can the AI take an action. It's should it, and can we prove it after the fact."
FAQ
Questions enterprises ask before engaging
Short answers to the questions that come up most often from boards, CISOs and AI platform teams scoping their first agentic security engagement.
What is Agentic Security and why does it matter now?
Agentic Security is the discipline of securing autonomous AI agents and the infrastructure they call into. Unlike a chatbot, an agent reasons, plans, invokes tools, spawns sub agents and chains decisions across systems. Each tool call is an executable action with real world impact, which means the security perimeter must move from the user session to the tool invocation itself. With the rapid adoption of Model Context Protocol, LangGraph, LangChain and frameworks like Azure AI Foundry, enterprises now ship agents into production faster than their security teams can catch up. Agentic Security is what closes that gap.
How does Agentic Security differ from traditional AI or LLM security?
Traditional LLM security focuses on the prompt and the model output: jailbreaks, prompt injection, hallucinations, data leakage at inference time. Agentic Security extends that perimeter to everything the agent can do once it has produced a plan, including authorization of tool calls, isolation of sub agents, observability of multi step workflows, reversibility of actions, and red teaming against tool chain abuse and data exfiltration. The blast radius is no longer a single response but an entire workflow.
What engagement models do you offer?
Three primary engagement models. Strategy: a board level diagnostic of your AI agent estate with a target architecture and a twelve month roadmap. Build: forward deployed engineering with your teams to implement guardrails, sub agent isolation, tool authorization and Langfuse observability in production. Assure: an independent agentic security audit including red teaming and a regulatory mapping pack ready for boardroom review. Fractional CISO engagements are also available for emerging platforms.
Which industries do you serve?
Regulated industries with high stakes, including banking, insurance and reinsurance, asset management, financial information providers, defense, healthcare, and the public sector. Twelve years of engagements span Europe, the GCC and Asia, with hands on experience for Swiss Re, AXA, BNP Paribas, Standard Chartered, Thomson Reuters, Société Générale, Seddiqi Holding, NusaVest, Thales, Dassault Systèmes, Veolia, Baxter and Gemalto.
Which frameworks and regulations do you align with?
ISO/IEC 42001 for AI management systems, NIST AI Risk Management Framework, and the EU AI Act for AI specific governance. ISO/IEC 27001 and 27701, SOC 2 Type II, HIPAA, PCI DSS, DORA, NIS2, and regional regimes such as UAE PDPL, KSA PDPL and GDPR for information security and data protection. Engagements consistently translate these frameworks into deployable controls rather than paperwork.
Where are you based and how do you work?
Based across Singapore, Dubai, Paris and Palo Alto, with active engagements across three continents. Most work is delivered remotely, with on site visits scoped to the engagement. Typical projects run from a few weeks for a strategy or audit, to multi quarter programmes for full builds and fractional CISO mandates.
How do we get started?
A short scoping call, typically thirty minutes, no commitment. The call clarifies your agent estate, your regulatory exposure and the outcome you need. Most engagements start with a structured diagnostic and a written proposal within ten working days.
Let's talk
Securing your agentic platform starts with a conversation.
Boards, CISOs, insurers and AI platform teams, if you're shipping autonomous agents, multi-cloud workloads, or building cyber-underwriting capability, I can help.
- Singapore
- Dubai
- Paris
- Palo Alto