MetaQuotes’ build 6140 release added more AI Assistant integrations and related controls to MetaTrader 5. Brokers should treat the feature as a workflow and vendor-governance change—not assume what user data an external model receives.
Follow the evidence
Trace how the event could reach markets, then inspect a competing explanation.
Compare explanations
Switch lenses to see what each account explains—and what remains uncertain.
The feature may help users navigate platform functions when prompts and permissions stay within a reviewed scope.
The feature may help users navigate platform functions when prompts and permissions stay within a reviewed scope.
Unclear key handling or prompt retention can create operational risk, so some firms may choose limited access or no rollout.
The release adds model-provider support and account-side controls
The MetaTrader 5 build 6140 notes describe expanded AI Assistant support for multiple language-model providers, with users able to enter API keys. The release also adds a way to delete chat histories and lists changes to the web terminal. MetaTrader 5: build 6140 release notes, 20 August 2026
The release notes establish product features; they do not specify how a particular broker configures access, which prompts are sent to a chosen provider or what contract applies to an account. Avoid assuming client credentials or trading data are shared unless the implementation and provider terms demonstrate that.
Build numbers identify a specific release, but broker terminals can update on different schedules and use different plugins. A feature visible in the vendor’s notes may therefore not match what a customer sees in a broker’s live environment.
The MetaTrader notes describe product capabilities but do not document a broker’s configuration or the chosen model provider’s data-retention terms. Those need to be checked in the service and privacy documentation for the actual deployment. MetaTrader 5: build 6140 release notes, 20 August 2026
AI in a trading terminal introduces a data-governance workflow
A broker considering rollout should map where keys are stored, which roles can invoke the assistant, what content is transmitted, how logs are retained and how a user can revoke access. Those controls are separate from whether the model’s answers are accurate.
Support teams may also need clear product boundaries: an assistant can help explain interface functions, but it should not be represented as a source of guaranteed market forecasts. Training, disclosure and human escalation paths help prevent a convenience feature from becoming unreviewed advice.
An AI assistant can help explain menus or platform functions, but a confident answer can still be wrong or outdated. A broker should decide which topics are allowed, test responses against support documentation and give customers a clear route to human help.
A broker’s testing should include the complete user journey: where the assistant is opened, which provider processes a prompt, how a key is entered or revoked and whether chat history deletion is visible to the user. Product notes are a starting point. The live broker configuration and provider terms determine the real data path.
Test access, retention and failure modes before launch
Ask whether API keys are held locally or by the platform, whether prompts may contain account identifiers, how model-provider retention works and what happens if an endpoint fails. Validate these answers in the exact build and broker configuration that users will see.
An alternative deployment choice is to keep the feature disabled or limited to educational help until governance is tested. The release itself does not require brokers to enable the assistant; its value depends on the controls, provider agreement and intended use.
A safe rollout can start with non-account-specific questions and a small internal test group. Record failure cases, verify how chat deletion behaves and review whether credentials can be rotated without interrupting ordinary platform access.