← Voltar ao feed
low

Proton launches Lumo Desktop agentic AI app with Connectors incl. shell access and per-action permissions

1 fonte(s) · 1 atualização(ões) · publicado 17/11/2026 09:54 · coletado há 5 horas

Proton has launched Lumo Desktop, a Windows, macOS and Linux app that turns its privacy-focused Lumo AI assistant into an agent. Its three new features are Connectors, Schedules and Skills. This is a product announcement, not a vulnerability or an incident. No CVEs, exploitation or threat actors are reported. The article says the source code is being published. It does not say whether an independent security review has been done. The research step was not run, so the record rests on the article alone and no Proton advisory, repository or release page was checked.

Detalhes técnicos

  • Connectors include local filesystem access, shell execution and browser content fetching, plus Proton Mail via Proton Mail Bridge, Dropbox, Jira, Confluence and custom local or remote MCP servers.
  • For each connector, users set which actions run automatically, which need approval and which are blocked. Proton's example is an assistant allowed to search an inbox, required to ask before sending an email and barred from deleting messages.
  • Connected tools are enabled in conversations by default, and users can disable them per conversation. The article notes that this default and the shell execution connector matter for managed endpoints.
  • Schedules run unattended tasks, and each task has its own permission set.
  • Proton says chat history uses zero-access encryption, Lumo keeps no server-side logs, and data is not used for model training. It says the data is hosted in Europe under Swiss law.

Remediação

Security teams assessing the app for managed endpoints should review the default-on connectors, especially shell and filesystem access. They should also restrict permissions to the minimum needed and require approval for sensitive actions.

Linha do tempo

  1. itsecurityguru.org ↗ · pub. 17/11/2026 09:54 · coletado 17/11 11:00 primeiro registro

    Proton launches Lumo Desktop agentic AI app with Connectors incl. shell access and per-action permissions

    ler a análise desta fonte

    Proton has launched Lumo Desktop, a Windows, macOS and Linux app that turns its privacy-focused Lumo AI assistant into an agent. Its three new features are Connectors, Schedules and Skills. This is a product announcement, not a vulnerability or an incident. No CVEs, exploitation or threat actors are reported. The article says the source code is being published. It does not say whether an independent security review has been done. The research step was not run, so the record rests on the article alone and no Proton advisory, repository or release page was checked.

    Technical Details

    • Connectors include local filesystem access, shell execution and browser content fetching, plus Proton Mail via Proton Mail Bridge, Dropbox, Jira, Confluence and custom local or remote MCP servers.
    • For each connector, users set which actions run automatically, which need approval and which are blocked. Proton's example is an assistant allowed to search an inbox, required to ask before sending an email and barred from deleting messages.
    • Connected tools are enabled in conversations by default, and users can disable them per conversation. The article notes that this default and the shell execution connector matter for managed endpoints.
    • Schedules run unattended tasks, and each task has its own permission set.
    • Proton says chat history uses zero-access encryption, Lumo keeps no server-side logs, and data is not used for model training. It says the data is hosted in Europe under Swiss law.

    Remediation

    Security teams assessing the app for managed endpoints should review the default-on connectors, especially shell and filesystem access. They should also restrict permissions to the minimum needed and require approval for sensitive actions.