For firms supervised by the AMF
Control how your firm uses AI, and show the AMF that you do.
The AMF has put AI at the heart of its 2026 strategy and has asked the firms it supervises to strengthen their defences. Qastel finds the AI in use across your firm, applies your policies where people work, and keeps a record of every decision.
See it in action
What the AMF asks, and how Qastel answers.
Five of the AMF's expectations, five answers in the product, in 42 seconds.
Regulatory context
AI is spreading. The AMF is paying closer attention.
- 90%
- use AI or plan to within a year
- 72%
- have an AI governance policy
- 83%
- of use cases are internal
AMF study of 2 February 2026, 100 financial market participants.
January
AI in the supervisory priorities
February
A study of AI use
June
A call to strengthen cyber defences
The challenge: move from an AI policy on paper to controls you can prove.
The risks
Your people use AI. Is your data protected?
Confidential data in a personal AI
A client portfolio pasted into a personal ChatGPT account.
With Qastel : Detected, redacted or blocked before it is sent.
AI tools nobody approved
New assistants adopted without IT or compliance knowing.
With Qastel : Every tool found, with its users and its risk.
Agents with too little oversight
An agent opens files or acts with no one checking.
With Qastel : Rules on access and actions, with human approval where supported.
The AMF's priorities
Each governance priority, one Qastel control.
Map systems and providers
AI inventory, shadow AI included
Protect confidential data
Warn, redact or block
Control access and actions
Rules for data and agents
Detect risky events
Alerts and a decision log
Strengthen internal control
Risk assessments and evidence
Raise staff awareness
A warning at the moment of risk
An illustration of how Qastel contributes, not a list of obligations set by the AMF.
The platform
Discover, assess, control, prove.
Discover
The AI in use
- Approved or not
- Personal accounts
- Agents and MCP
Assess
Your risks, ranked
- Confidentiality
- Providers
- Agent permissions
Control
Your policy, applied
- “Block client data to personal AI accounts.”
- “Require approval for a sensitive agent action.”
- Test in monitoring mode first.
Prove
The evidence, kept
- Violations logged
- Decisions traced
- Ready for audit
Built for financial services
Independent, European, ready for your infrastructure.
- Independent of model providers
- Built in Paris, hosted in Europe
- Prompts analysed without being kept
- Connects to your identity and security tools
- Roll out gradually, in monitoring first
- Shared by business, compliance, risk and IT
Questions from compliance and risk officers
Does the AMF require an AI governance platform?
No. The AMF does not prescribe Qastel or any platform. Qastel helps you put in place and document controls suited to your AI use.
Does Qastel guarantee compliance with DORA or the EU AI Act?
No. Qastel helps you identify and control risks and keep evidence. Compliance takes an overall programme suited to your obligations.
Does Qastel keep what people type?
Prompt analysis is off by default. When an administrator turns it on, prompts are classified in memory without the text being kept.
What can we show in an inspection?
The inventory of AI systems, their risk assessments, the policies in place and the record of how they were applied.
You have an AI policy. Is it really applied?
In 30 minutes, see how Qastel finds your AI use, applies your rules and documents your controls.
Sources
- AMF: supervisory priorities for 2026, 14 January 2026
- AMF: study on the use of AI, 2 February 2026
- AMF: cyber resilience and AI, 3 June 2026
This page presents AMF publications for information. Qastel is not affiliated with or endorsed by the AMF. The platform alone does not guarantee regulatory compliance.
