OpenAI's dots Assistant Exposes Permission and Financial Transaction Gaps

What happened

Why it matters

• ·Employees who connect personal or work accounts to dots will expose email, calendar, and file content to OpenAI's platform continuously, not just during discrete sessions. This creates persistent data flows that existing employee AI-use policies and data classification controls may not cover.
• ·The absence of read-only access options means enterprise data stored in connected cloud drives is subject to write permissions by default. This conflicts directly with least-privilege data access principles required under frameworks such as the EU AI Act (Regulation (EU) 2024/1689) and expected under most enterprise data governance standards.
• ·OpenAI's planned expansion into agent-initiated financial transactions echoes the unresolved spending authority gap flagged in Agents With Spending Authority Need Behavioral Testing First, Research Warns. Compliance teams that have not yet established pre-authorization controls for agentic spending face compounding exposure as this capability moves closer to release.

Governance controls affected

What to do now

• ☐Survey employees now to determine whether any have already connected work email, calendar, or cloud storage accounts to ChatGPT or dots, and require disconnection pending a formal review.
• ☐Update your acceptable-use policy for AI tools to explicitly address always-on assistants that maintain persistent access to communication and file systems, not just tools used on demand.
• ☐Ask your IT and procurement teams to confirm whether your cloud storage and email systems can enforce read-only access when third-party AI tools request connection, and document any gaps.
• ☐Establish a written policy requiring pre-approval before any AI tool is permitted to initiate financial transactions on behalf of employees or the organization, and confirm this policy covers consumer-facing AI products employees might use personally.
• ☐Add dots and similar persistent-access AI assistants to your shadow AI inventory process and flag them for a data protection impact assessment before any sanctioned enterprise use.

What to watch next

Compliance teams should monitor OpenAI's timeline for expanding permission granularity in dots, and track whether payment processor negotiations result in a live transaction capability. Any such capability launch would require an immediate review of spending authority controls. It may also trigger notification obligations under data protection laws, particularly for organizations with EU operations under the EU AI Act (Regulation (EU) 2024/1689). Regulators already probing agentic AI data practices, including the UK's Information Commissioner's Office following ICO Secures Developer Changes and Opens Agentic AI Data Protection Probe, may treat always-on personal assistants as a priority enforcement category.